Flood Resilience Rose Across 92 Chinese Cities, Study Finds

Cities are getting better at living with floodwater, and the improvement isn't just in the richest ones. A study published in Scientific Reports on November 14, 2026, assessed 92 cities in five Chinese urban agglomerations over 2006 to 2022, scoring urban flood resilience through 19 indicators across four dimensions. Mean resilience rose from 0.443 to 0.516 across the period.

The gains also spread out. The inter-city Gini coefficient declined by 19.29 percent, which means overall improvement coincided with narrowing relative disparities between cities, although the authors note that gains still differed among the five agglomerations. Resilience rose as a regional condition, not only as a headline number in a few large metros.

The diagnosis of what's missing is specific. Local supporting conditions showed prominent relative shortfalls in flood-protection and drainage indicators, the pipes-and-barriers layer of resilience, and regional differences in vegetation-related constraints. A city can score well on response and recovery and still be held back by the drainage it hasn't built.

Geography carried unusual weight in the modeling. The team combined spatial analysis, inequality decomposition and obstacle diagnosis with interpretable machine learning, and its SHAP analysis of an XGBoost model found that mean slope and population density together accounted for 51.04 percent of mean absolute SHAP contributions. Where a city sits, and how tightly its people are packed, shaped its score more than any single program.

The forward look tested five development scenarios. The coordinated development scenario produced the highest projected flood resilience, while the population agglomeration scenario showed only a small additional gain over business as usual. Piling more people into the same urban footprint, on its own, buys almost nothing.

The authors, led by Zhen Li and Qinlong Hu of Chengdu University of Technology, frame the findings as a basis for agglomeration-level resilience plans that account for local conditions and address flood-control infrastructure and ecological capacity together. The paper was received on August 26, 2026, and accepted on October 31, 2026.

The method is part of the result's weight. Resilience wasn't scored on flood damage alone; the 19 indicators span the capacities that support a city before, during and after water arrives, and the five scenarios were modeled rather than asserted. That makes the coordinated-development finding a planning claim with numbers behind it, and it gives other fast-urbanizing regions a template: measure the drainage shortfall city by city, because the regional average hides exactly the gaps that flood first.