Substantial increase in future fluvial flood risk projected in China's major urban agglomerations

Floods cause substantial global losses and casualties and are expected to intensify with ongoing warming that accelerates the hydrological cycle. Most projections of future flood hazard have used CMIP5-based outputs; however, CMIP6 provides improved precipitation simulations. China accounts for about 10% of global flood losses and faces heightened flood risk amid rapid urbanization, with urban land in floodplains having expanded over 500% from 1992 to 2015. Existing studies have examined flood risk with urban growth or climate change, but often omit urban spatial expansion or rely on older climate scenarios. The relative roles of climate change versus urban expansion in driving urban flood risk, and the effects of expansion patterns, remain unclear. Leveraging bias-corrected CMIP6 forcings and new high-resolution projections of urban land expansion under the SSP framework, this study assesses future fluvial flood risk across seven major Chinese urban agglomerations, quantifying contributions from climate change and urban expansion and examining how expansion patterns affect inundation exposure.

Prior work has primarily used CMIP5-driven hydrological models for flood projections; CMIP6 shows improved precipitation extremes over China. Studies have linked socio-economic development to increased exposure, noting rapid urbanization in China and recent catastrophic events (e.g., Zhengzhou 2021). Some research incorporated urban growth but not climate change, while others included climate change but modeled only densification within existing urban cells, omitting spatial expansion. Muis et al. considered both but for 2010–2049 and without SSP-based urban expansion. The literature indicates knowledge gaps regarding longer-term dynamics post-2050, SSP-consistent urban expansion, and how spatial patterns of urban growth (e.g., edge expansion) modulate flood risk. New 1-km SSP-based urban expansion datasets (Chen et al. 2020) and higher-resolution flood maps now enable integrated assessment of these factors.

Forcings and scenarios: Five ISIMIP3b bias-adjusted CMIP6 GCMs (GFDL-ESM4, IPSL-CM6A-LA, MPI-ESM1-2-HR, MRI-ESM2-0, UKESM1-0-LL) provided daily max/min temperature, precipitation, and wind (0.25° after bilinear interpolation) for historical and SSP1-2.6, SSP3-7.0, SSP5-8.5 scenarios. Analyses compare the 1985–2014 reference to 30-year windows centered on when global mean temperatures reach 1.5, 2.0, 2.5, and 3.0 °C above preindustrial, per ISIMIP3b temperature thresholds. Under SSP1-2.6, only 1.5 °C is considered (insufficient GCMs reach higher levels). Model framework: The VIC hydrological model simulated daily runoff at 0.25° over China using ISIMIP3b and CMFD forcings; soil/vegetation parameters followed prior calibrated studies. Runoff forced the CaMa-Flood v3.6 hydrodynamic model to compute daily discharge, river water storage, flood depth, and inundation at 0.25°. River width used GWD-LR; river depth was calibrated. Inundation downscaling mapped 0.25° water levels to an 18-arcsec (~500 m) DEM by sub-grid elevation thresholding, conserving water volume, a standard approach in flood risk studies. Flood protection and frequency: Given inconsistencies and coarse resolution in available protection datasets for China, the study assumed consideration of events exceeding historical 100-year floods. For each 0.25° cell, annual maximum daily river water storage (30-year historical) was fit with a Gumbel distribution to estimate the 100-year magnitude; arid areas with average runoff <0.02 mm/day (from CMFD-forced VIC) were excluded. CMFD-based flood depths were used to establish depth–recurrence relationships, transforming GCM-simulated recurrence to consistent depths. A sensitivity test with a three-parameter Weibull distribution yielded comparable exposure increases. Study regions: Seven major urban agglomerations were analyzed: Beijing–Tianjin–Hebei (BTH), Central Plains (CP), Central Shanxi Plain (CSP), Yangtze River Delta (YRD), Triangle of Central China (TCC), Chengdu–Chongqing (CC), Pearl River Delta (PRD), and their aggregate. Urban land data: Future urban expansion followed Chen et al. (2020) 1-km SSP-based projections (FLUS cellular automata with ANN-estimated development potential, neighborhood effects, constraints), resampled to 18-arcsec via nearest neighbor to align with inundation maps. Historical reference year for urban land is 2015; future uses the closest decade to each warming window. Variables tracked included total area, urban/non-urban areas and inundated areas, and inundated proportions of total, urban, and non-urban land. Decomposition of drivers: Three assumptions quantified contributions: (i) SSP (socio-economic/urban expansion only; climate hazard unchanged), (ii) RCP (climate change only; urban land fixed at historical), and (iii) SSP-RCP (both changing). Inundated urban land areas were computed by overlaying urban maps with inundation maps. Contributions of climate change and urban expansion were calculated from relative changes under SSP and RCP vs historical, normalized by absolute sums. Additional analyses: HAND from MERIT Hydro characterized proximity to drainage to examine spatial patterns of new urban expansion relative to rivers and historical 100-year floodplains. Ensemble handling: Unless noted, results report ensemble medians across the five GCMs, with uncertainty summarized by min–max or quantiles. Model performance and uncertainties were assessed in supplementary notes.

- Urban inundation increases: Across seven agglomerations, 30-year average annual inundated urban land areas increase by factors of 4.7–8.3 (1.5 °C), 4.5–9.7 (2.0 °C), 10.8–12.8 (2.5 °C), and 15.5–19.0 (3.0 °C) relative to historical, depending on scenario. These correspond to absolute increases of roughly 110–190, 100–220, 250–290, and 350–430 km², respectively. - Regional contrasts: PRD, CC, and TCC show the largest increases; YRD maintains large absolute inundated areas due to large historical exposure. Flood risk does not increase monotonically in all cases and is generally highest under SSP5-8.5. - Inundated proportions: Historical average annual inundated proportion of urban land is 0.04% (aggregate). Under warming, it rises to 0.20–0.32% (1.5 °C), 0.20–0.36% (2.0 °C), 0.42–0.44% (2.5 °C), and 0.59–0.63% (3.0 °C), i.e., 5–8x, 5–9x, 10–11x, and 15–16x increases. PRD’s annual urban inundated proportion exceeds 1% under all scenarios. - Urban more exposed than non-urban: Except in YRD, the urban inundated proportion exceeds that of total land by >20%, indicating urban areas are more prone to flooding (often near rivers). - Driver attribution: Climate change is the dominant driver (>50% contribution in most cases), reaching near 100% in southern China; urban expansion plays a larger role in some northern agglomerations (e.g., BTH, CSP) and at lower warming levels (<2 °C). As warming increases and urban expansion slows after ~2040–2050, the relative contribution of expansion declines. - Effect of urban expansion: Including both climate change and urban expansion yields 10–50% higher inundated urban areas than climate change alone (SSP-RCP vs RCP) under SSP3-7.0 and SSP5-8.5. - Spatial pattern matters: Edge-expansion causes new urban development to occur closer to rivers, increasing exposure. Newly-developed urban land has, on average, about 10% greater inundated extent (as percent of new urban area) than historical urban land. Exceptions include CC (mountainous) and PRD (already saturated river-adjacent urbanization), where new development shifts to higher terrain further from rivers; YRD’s flat terrain and large floodplain make it less sensitive to HAND changes. - Hazard intensification: The historical 100-year flood becomes more frequent over most of China (shorter return periods), especially in the south, consistent with increased extreme precipitation. - Adaptation implications: Historical 100-year protection reduced inundated urban land by >90%, but will not suffice under future risk; policy already recommends 200-year or higher protection for large, wealthy cities; integrated strategies are needed (mitigation, urban planning to reduce exposure, and strengthened defenses).

The study shows a pronounced rise in urban fluvial flood risk in China’s major agglomerations driven primarily by climate change, with urban spatial expansion further elevating risk by 10–50% beyond climate effects alone. Southern China exhibits particularly strong climate-driven increases due to intensified extreme precipitation and more frequent historical 100-year events. In northern agglomerations, urban expansion contributes relatively more at lower warming levels and where climate intensification is weaker. The spatial configuration of growth—dominated by edge expansion—places new development closer to rivers and historical floodplains, making it more flood-prone than existing urban areas, except where terrain or prior saturation push growth to higher ground. These findings underscore the need to limit warming below 2 °C (preferably 1.5 °C), account for urban expansion in risk assessments and loss projections, and plan urban growth to avoid flood-prone zones. Given that 100-year protection will be inadequate under future hazards, adaptation requires higher protection standards alongside exposure and vulnerability reductions. Despite uncertainties across forcings, models, and datasets, ensemble agreement on increasing inundated urban areas is high in most regions, and sensitivity analyses confirm robustness of the qualitative conclusions.

Using bias-corrected CMIP6 forcings and high-resolution SSP-based urban expansion data, the study projects that inundated urban land in seven major Chinese agglomerations will increase by roughly 4–19 times under 1.5–3.0 °C warming. Climate change is the predominant driver, but urban spatial expansion significantly amplifies risk, particularly through edge-expansion that places new urban land closer to rivers; newly-developed urban land is, on average, more flood-prone than historical urban land. The results highlight the urgency of integrating climate mitigation, risk-aware urban planning that limits expansion into flood zones, and upgraded flood protection (beyond 100-year) to minimize future losses. Future research should incorporate human interventions (e.g., reservoirs, urban drainage), urbanization-induced hydrologic changes, risk-aware urban growth modeling under SSPs, and extend analyses to other agglomerations to generalize findings.

Key limitations include: (1) Exclusion of human interventions such as reservoirs, water withdrawals, and urban drainage networks, potentially overestimating inundation; (2) Urban expansion projections (FLUS) did not account for flood risk, so some newly-developed urban land falls within historical 100-year floodplains, potentially inflating projected exposure; (3) Urbanization-induced changes in hydrology (e.g., imperviousness increasing runoff) were not represented, though these effects are strongest in small catchments; (4) Uncertainties in climate forcings, hydrologic/hydrodynamic models, and flood frequency fitting persist, although ensemble analyses and alternative distributions indicate robust qualitative outcomes; (5) The focus on seven agglomerations excludes other Chinese regions that may face similar risks; (6) Assumed 100-year protection benchmark in historical reference may differ locally.