Li et al. (2026) Quantifying rainfall variability and potential hazards of extreme events in Beijing through stochastic simulations

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-01-06
• Authors: T. D. Li, Shuiqing Yin, Yuanyuan Xiao, Maoqing Wang, Liutong Chen, Wenyue Zou, Nadav Peleg
• DOI: 10.1016/j.ejrh.2025.103068

Research Groups

• State Key Laboratory of Earth Surface Processes and Hazards Risk Governance (ESPHR), Faculty of Geographical Science, Beijing Normal University, Beijing, China
• Institute of Earth Surface Dynamics, University of Lausanne, Lausanne, Switzerland
• Expertise Center for Climate Extremes, University of Lausanne, Lausanne, Switzerland

Short Summary

This study quantifies rainfall variability and potential hazards of extreme events in Beijing using stochastic simulations. It reveals that rainfall variability increases with finer spatiotemporal resolution and longer return periods, with 100-year return period rainfall potentially exceeding station records by up to 53%, highlighting significant underestimation of extreme event hazards from short observational records.

Objective

• To quantify natural rainfall variability and the "true" potential range of extreme rainfall in Beijing.
• To enhance the ability to capture natural rainfall variability for more robust estimates of extreme rainfall.
• To generate multiple realizations of 30-year rainfall fields at 1 km resolution and hourly intervals using a stochastic 2-dimensional rainfall generator.
• To advance the assessment of extreme rainfall hazards through event-based extreme value analysis.
• To provide 10–100-year extreme rainfall return-level estimates and plausible ranges for urban drainage design and flood-risk evaluation.

Study Configuration

• Spatial Scale: Beijing metropolitan area, China. Gridded rainfall simulations at 1 km resolution, covering a domain from 1 km (grid scale) to 70 km (averaged domain scale).
• Temporal Scale: Hourly intervals for simulations and observations. Stochastic simulations generated 30 realizations of 30 years each. Observational data spanned 1991–2020 for weather stations and 2015–2020 for gridded rainfall. Event-based extreme value analysis was performed.

Methodology and Data

• Models used:
  • AWE-GEN-2d (Advanced WEather GENerator for a two-dimensional grid) for stochastic rainfall simulation.
  • Weibull distribution for fitting annual maxima series in event-based Intensity-Duration-Frequency (IDF) curve calculation.
• Data sources:
  • Hourly rainfall data from 10 weather stations (China Meteorological Administration, 1991–2020).
  • Hourly gauge-satellite-merged gridded rainfall data: CMPAS (China Multisource Precipitation Analysis System) at 1 km resolution (2015–2020).
  • Satellite-based rainfall estimates: CMORPH (U.S. Climate Prediction Center, 30-minute temporal, 8 km spatial resolution).
  • Radar quantitative rainfall observations (Meteorological Detection Center of CMA, 1 hour temporal, 0.01° spatial resolution).
  • Latitudinal (U) and longitudinal (V) wind data at 850 hPa from MERRA-2 climate reanalysis product (0.5° × 0.625° spatial, hourly temporal resolution, 2011–2019).

Main Results

• Rainfall variability (V) increases significantly with finer temporal scales, peaking at the event-based level (V > 1.5 for all months at event scale).
• V also increases as the spatial scale becomes finer, from the averaged domain scale (70 km) to the grid scale (1 km).
• For extreme events, variability grows markedly with longer return periods.
• Observed event-based extreme rainfall return levels were 172 mm (10-year), 232 mm (50-year), and 255 mm (100-year).
• Stochastic simulations (95th percentile) yielded significantly higher extreme rainfall return levels: 196 mm (10-year), 329 mm (50-year), and 391 mm (100-year).
• The 100-year return period rainfall from simulations can exceed station records by up to 53% (391 mm vs 255 mm), indicating substantial hazard underestimation by short records.
• The highest rainfall return levels are primarily concentrated southwest of Beijing.
• Upper-bound scenarios (P95) show areal average rainfall return levels 1.4 to 1.6 times higher than the median (and observations) for 10- to 100-year return periods.
• For the 100-year return period, 40% of grid values in the P95 scenario exceed 500 mm, whereas nearly all grid values in the median scenario remain below 500 mm.

Contributions

• Proposes and validates a gridded rainfall generator (AWE-GEN-2d) to extend limited rainfall records and simulate yet-unobserved extreme rainfall events, effectively capturing natural climate variability.
• Emphasizes the heterogeneity of rainfall fields and quantifies the hazards of plausible, yet-unobserved, extreme rainfall events under a stationary climate.
• Provides critical insights for improving urban flood management strategies and long-term infrastructure planning in highly urbanized regions like Beijing.
• Advances extreme rainfall hazard assessment by employing an event-based extreme value analysis, which better reflects the cumulative impact of full storm systems compared to fixed-duration approaches.
• Demonstrates the effectiveness of stochastic rainfall generators in improving the estimation of return levels for extreme rainfall, even when calibrated with relatively short observational records.
• The developed framework is broadly applicable to other regions and hydrological studies beyond urban settings.

Funding

• National Key Research and Development Program of China (Grant No. 2020YFF0304401)
• National Key Research and Development Program of China (Grant No. 2021YFE0113800)
• China Scholarship Council (CSC Grant number: No. 202406040216)
• China Scholarship Council (CSC Grant number: 202106040028)
• Switzerland National Science Foundation (SNSF Grant number: 194649, "Rainfall and floods in future cities")

Citation

@article{Li2026Quantifying,
  author = {Li, T. D. and Yin, Shuiqing and Xiao, Yuanyuan and Wang, Maoqing and Chen, Liutong and Zou, Wenyue and Peleg, Nadav},
  title = {Quantifying rainfall variability and potential hazards of extreme events in Beijing through stochastic simulations},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2025.103068},
  url = {https://doi.org/10.1016/j.ejrh.2025.103068}
}