Data-Driven Modelling of Structural Interdependencies in Urban Flood-Resilience Support Systems: A Panel Granger Network Approach

Urban flood resilience depends on interactions among social protection, ecological conditions, economic resources, and infrastructure services. This study develops a panel Granger network framework to examine temporal predictive relationships among 15 resilience-support indicators in 13 cities in the Beijing–Tianjin–Hebei region during 2011–2022. Annual series were directionally harmonized, first-differenced, and standardized. One-lag Dumitrescu–Hurlin tests, followed by nominal Benjamini–Hochberg screening, identified 26 candidate directed links among 210 ordered indicator pairs; signed edge weights were estimated using two-way fixed effects. Of the retained links, 22 (84.62%) connect different dimensions, and the network contains three mixed-dimension communities within one weakly connected component. A four-metric Borda ranking places the total age-dependency ratio, medical-insurance enrolment ratio, R&D expenditure relative to GDP, and wastewater-treatment rate in the first four positions; drainage-pipe density and health technicians share fifth place. Leave-one-city-out analysis retains 80.8–96.2% of baseline links, compared with 23.1–57.7% under leave-one-year-out analysis. Targeted removal experiments further characterize changes in directed reachability and network connectivity. These findings establish an empirical description of cross-dimensional predictive organization and reveal the sensitivity of the selected network to temporal sample composition. The framework integrates panel-based link screening, network characterization, and specification diagnostics for the structural analysis of urban flood-resilience support indicators.