Weng et al. (2025) Spatially coherent changes in Chinese annual flood peaks revealed by a consensus-based machine learning framework for regionalization

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-26
• Authors: J. Weng, Yixin Yang, Dayang Li, Ashish Sharma, Long Yang
• DOI: 10.1016/j.jhydrol.2025.134665

Research Groups

• School of Geography and Ocean Science, Nanjing University, Nanjing, China
• Frontiers Science Center for Critical Earth Material Cycling, Nanjing University, Nanjing, China
• College of Civil Engineering, Yancheng Institute of Technology, Yancheng, China
• School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia

Short Summary

This study develops a consensus-based machine learning framework to identify homogeneous flood regions across China, revealing predominant trends of decreasing annual flood peak magnitudes and delayed occurrences in most regions, primarily driven by climate factors.

Objective

• To develop a consensus-based machine learning framework that objectively identifies homogeneous flood regions by combining hierarchical and ensemble clustering while mitigating descriptor-dependent biases.
• To apply this framework to analyze spatially coherent changes in annual flood peak magnitudes and timings across China and identify their primary drivers.

Study Configuration

• Spatial Scale: China, utilizing data from 1111 stream gauging stations.
• Temporal Scale: 1980–2017.

Methodology and Data

• Models used: Consensus-based machine learning framework, which integrates hierarchical and ensemble clustering techniques.
• Data sources: Continuous observations of annual flood peaks (including peak magnitude and timing) from 1111 stream gauging stations across China.

Main Results

• Twenty homogeneous flood regions were identified across China, each exhibiting similar flood regimes distinct from others.
• Indices characterizing the climatologically mean flood regime (e.g., mean normalized flood discharge) were found to be more influential in flood regionalization than those representing interannual variability (e.g., coefficient of variation of flood discharge).
• Regional aggregation demonstrated stronger temporal persistence in annual flood peak series compared to individual stations.
• Predominant trends of decreasing flood peak magnitudes were observed in 15 out of 20 regions, while delayed occurrences were noted in 15 out of 20 regions.
• These spatially coherent flood changes are primarily climate-driven, with dominant controls varying regionally among rainfall, snowmelt, and soil moisture dynamics.

Contributions

• Developed a novel and transferable consensus-based machine learning framework for objective flood regionalization that mitigates descriptor-dependent biases.
• Provided the first comprehensive national-scale assessment of Chinese flood regimes and their spatially coherent changes.
• Offered critical insights for adaptive flood management and a robust framework for climate impact analysis.

Funding

• Not specified in the provided text.

Citation

@article{Weng2025Spatially,
  author = {Weng, J. and Yang, Yixin and Li, Dayang and Sharma, Ashish and Yang, Long},
  title = {Spatially coherent changes in Chinese annual flood peaks revealed by a consensus-based machine learning framework for regionalization},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134665},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134665}
}