Guan et al. (2025) The ability of a stochastic regional weather generator to reproduce heavy-precipitation events across scales

Identification

• Journal: Natural hazards and earth system sciences
• Year: 2025
• Date: 2025-09-05
• Authors: Xiaoxiang Guan, Nguyễn Viết Dũng, Paul Voit, Bruno Merz, Maik Heistermann, Sergiy Vorogushyn
• DOI: 10.5194/nhess-25-3075-2025

Research Groups

• GFZ Helmholtz Centre for Geosciences, Section Hydrology, Potsdam, Germany
• Institute of Environmental Sciences and Geography, University of Potsdam, Potsdam, Germany

Short Summary

This study assesses a non-stationary regional weather generator's (nsRWG) ability to reproduce heavy-precipitation event (HPE) extremity across spatial and temporal scales in Germany, finding it largely excels in replicating observed extremity patterns and potential influential areas, making it suitable for flood risk assessment.

Objective

• To assess the ability of a regional weather generator to represent the extremity of heavy-precipitation events (HPEs) across spatial and temporal scales.
• To evaluate how well the cross-scale extremity of precipitation events is captured by a weather generator, even if it is not specifically tailored or trained to do so.

Study Configuration

• Spatial Scale: Germany and adjacent regions, covering the five major river catchments (Elbe, Rhine, Danube, Ems, Weser), totaling over 650,000 km².
• Temporal Scale: 72 years (1950–2021) for synthetic daily precipitation data, matching the length of observed data; precipitation durations from 1 to 7 days.

Methodology and Data

• Models used:
  • Multi-site non-stationary regional weather generator (nsRWG)
  • First-order multi-variate auto-regressive (MAR-1) model for spatiotemporal dependence
  • Type-1 extended generalized Pareto (extGP) distribution for daily non-zero precipitation amounts
  • SANDRA (Simulated ANnealing and Diversified RAndomization) based on k-means clustering for circulation pattern classification
  • Generalized Extreme Value (GEV) distribution, specifically duration-dependent GEV (dGEV) method, for return period estimation
• Data sources:
  • Gridded precipitation data: E-OBS (version 25.0e) at 0.25° grid, resampled to 0.5° resolution, daily, 1950–2021.
  • Mean sea level pressure and daily air temperature at 2 m height: ERA5 reanalysis, daily, 1950–2021.
  • Weather extremity index (WEI) and cross-scale weather extremity index (xWEI) for quantifying event extremity.

Main Results

• The nsRWG excels in replicating the extremity patterns for almost all seven durations (ranging from 1 to 7 days) considered.
• The frequency of small-scale 1-day rainfall is slightly overestimated by the nsRWG.
• The nsRWG aptly reproduces the potential influential areas of HPEs, whether of a short or long duration.
• It is capable of generating precipitation events mirroring the extremity patterns observed during past disaster-causing HPEs in Germany (e.g., August 2002 event), while simultaneously accommodating their variations.
• Simulated distribution patterns of annual maximum WEI and corresponding areas are in good agreement with E-OBS observations.
• For shorter durations (1 and 2 days), the frequencies of HPEs with WEI areas ≤20,000 km² are overestimated, while those with larger WEI areas are underestimated.
• Empirical probability plots of annual maximum WEI and xWEI show good agreement between observed and simulated data, particularly for high-return periods.
• The nsRWG tends to slightly underestimate observed WEI and xWEI for return periods between 2 and 5 years.

Contributions

• Introduces and demonstrates a novel evaluation approach for a weather generator based on the Weather Extremity Index (WEI) and the cross-scale Weather Extremity Index (xWEI) to assess its ability to capture cross-scale rainfall extremity.
• Provides a comprehensive evaluation of a multi-site stochastic regional weather generator (nsRWG) at a large spatial scale (covering all of Germany and riparian regions).
• Demonstrates the potential of the nsRWG for simulating HPE-related hazards and assessing flood risks, including the generation of synthetic events with spatiotemporal extremity patterns similar to historical disaster-causing HPEs.
• Highlights the utility of weather generators in exploring spatial counterfactual scenarios for extreme events.

Funding

• Project FLOOD (project no. 01LP2324E) as part of the ClimXtreme Research Network on Climate Change and Extreme Events within the framework program Research for Sustainable Development (FONA3).
• China Scholarship Council (grant no. 202106710029) for Xiaoxiang Guan's PhD research.
• Article processing charges covered by the GFZ Helmholtz Centre for Geosciences.

Citation

@article{Guan2025ability,
  author = {Guan, Xiaoxiang and Dũng, Nguyễn Viết and Voit, Paul and Merz, Bruno and Heistermann, Maik and Vorogushyn, Sergiy},
  title = {The ability of a stochastic regional weather generator to reproduce heavy-precipitation events across scales},
  journal = {Natural hazards and earth system sciences},
  year = {2025},
  doi = {10.5194/nhess-25-3075-2025},
  url = {https://doi.org/10.5194/nhess-25-3075-2025}
}