Wei et al. (2026) Integrating non-stationarity into extreme rainfall risk assessment: A GAMLSS-based framework for large-scale region

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-01-10
• Authors: Wanyin Wei, Lu Xia, Lanjun Li, Lu Xia
• DOI: 10.1016/j.jhydrol.2026.134934

Research Groups

• State Key Laboratory of Water Engineering Ecology and Environment in Arid Area, Xi’an University of Technology, Xi’an, China
• College of Resources and Environment, Shanxi Agricultural University, Taigu, China

Short Summary

This study developed a framework to integrate non-stationarity into extreme rainfall risk assessment, applying it to the Beijing-Tianjin-Hebei region of China. The research demonstrated that incorporating non-stationarity, using a GAMLSS-based approach, significantly alters the spatial patterns of assessed disaster risks compared to traditional stationary methods.

Objective

• To construct and apply a framework for incorporating non-stationarity into potential risk assessment of extreme rainfall in the Beijing-Tianjin-Hebei region of China, addressing the potential underestimation of risks when non-stationarity is neglected.

Study Configuration

• Spatial Scale: Beijing-Tianjin-Hebei region, China
• Temporal Scale: 1980 to 2023

Methodology and Data

• Models used: Generalized Additive Models for Location, Scale, and Shape (GAMLSS) framework. Specific distributions identified within GAMLSS include Gamma (for R10) and Log-Normal (for RX1, R95p, CWD).
• Data sources: Precipitation data used to derive four extreme precipitation indices (EPIs): maximum 1-day precipitation (RX1), annual count of days with precipitation ≥10 mm (R10), sum of daily precipitation amounts exceeding the 95th percentile (R95p), and maximum number of consecutive wet days (CWD). (Likely observational/gridded data, though not explicitly stated).

Main Results

• Extreme rainfall indices exhibited significant spatial clustering patterns across the study area and generally showed downward temporal trends.
• Within the GAMLSS framework, the Gamma distribution was recommended for R10, while the Log-Normal distribution was most suitable for RX1, R95p, and CWD.
• Non-stationarity models were identified as optimal GAMLSS models for a significant proportion of grids: 50.2 % for RX1, 79.8 % for R10, 62.7 % for R95p, and 71.6 % for CWD.
• Risk assessment incorporating non-stationarity revealed increasing disaster risks in northwest mountainous regions, contrasting with decreasing trends observed in the central study area, compared to assessments without non-stationarity.

Contributions

• Development and application of a novel framework for integrating non-stationarity into extreme rainfall risk assessment, considering it as a critical disaster-causing factor alongside environmental-inducing and disaster-bearing factors.
• Provides insights into the significant impact of accounting for non-stationarity on the spatial patterns of extreme rainfall disaster risks, particularly for large-scale regions.
• Offers a robust methodology for identifying appropriate probability distributions and optimal GAMLSS models for various extreme precipitation indices under non-stationary conditions.

Funding

• Not specified in the provided text.

Citation

@article{Wei2026Integrating,
  author = {Wei, Wanyin and Xia, Lu and Li, Lanjun and Xia, Lu},
  title = {Integrating non-stationarity into extreme rainfall risk assessment: A GAMLSS-based framework for large-scale region},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.134934},
  url = {https://doi.org/10.1016/j.jhydrol.2026.134934}
}