Costabile et al. (2026) A stochastic rain-on-grid framework for handling spatio-temporal rainfall uncertainty in impact-based flood nowcasting

Identification

• Journal: International Journal of Disaster Risk Reduction
• Year: 2026
• Authors: Pierfranco Costabile, Margherita Lombardo, Carmelina Costanzo, Ioannis Tsoukalas, Vasilis Bellos
• DOI: 10.1016/j.ijdrr.2026.105998

Research Groups

• Department of Environmental Engineering, University of Calabria, Rende, Italy.
• Department of Civil Engineering, Democritus University of Thrace, Xanthi, Greece.
• Department of Environmental Engineering, Democritus University of Thrace, Xanthi, Greece.

Short Summary

This study introduces a Stochastic Rain-on-Grid framework that couples high-resolution stochastic rainfall generation with 2D hydrodynamic modeling to quantify how rainfall spatio-temporal uncertainty propagates into flood impacts. The findings demonstrate that while hydrological responses are highly sensitive to rainfall structure, street-level hazard classifications are more robust, providing a more stable target for early warning systems.

Objective

• To develop an integrated, modular framework for impact-oriented flash flood nowcasting that explicitly accounts for the stochastic nature of rainfall's spatial and temporal organization.

Study Configuration

• Spatial Scale: Watershed scale (Agia Aikaterini and Soures catchments, ~50 km²) and street-level urban scale (Mandra town, Greece).
• Temporal Scale: Flash flood event duration (approx. 13 hours) with a data resolution of 2 minutes.

Methodology and Data

• Models used:
  • anySim: A stochastic rainfall generator used to produce 100 equiprobable storm realizations based on a zero-inflated exponentiated Weibull distribution and separable spatiotemporal correlation structures.
  • MYTHOS-2D (formerly UniCal): A 2D hydrodynamic model solving Shallow Water Equations (SWE) via the Finite Volume Method, including a Green-Ampt infiltration module.
• Data sources: High-resolution X-band dual-polarization Doppler weather radar (XPOL) data from the National Observatory of Athens (NOA) with a resolution of 200 m in space and 2 min in time.

Main Results

• Variability Attenuation: Hydrological metrics at the catchment scale showed high sensitivity to rainfall structure (Coefficient of Variation, CV, up to 0.53 for peak discharge), but this variability attenuated at the street scale (average CV ~0.25 for hazard classes).
• Predictive Metrics: The temporal coefficient of variation (Tcv) and spatial coefficient of variation (Scv) of rainfall were identified as the most consistent predictors of flood impact severity.
• Storm Variability Diagram: A new classification tool was proposed to group equiprobable events by expected impact; partitioning the ensemble based on a Tcv threshold increased spatial agreement in hazard mapping from 48% to 84%.
• Quantitative Impact: Peak discharges across the ensemble ranged from 14 m³/s to over 320 m³/s, despite statistically similar total rainfall volumes (CV < 10%).

Contributions

• Methodological Integration: Bridges the gap between stochastic meteorological modeling and high-resolution urban hydrodynamic impact assessment.
• Robustness of Impact Metrics: Provides evidence that hazard-based impact proxies (water depth and velocity classes) are more reliable for operational forecasting than traditional discharge-based metrics.
• Operational Tool: Introduces the "Storm Variability Diagram" as a proof-of-concept for rapid, uncertainty-aware screening of storm realizations in nowcasting systems.

Funding

• Not specified in the provided text.

Citation

@article{Costabile2026stochastic,
  author = {Costabile, Pierfranco and Lombardo, Margherita and Costanzo, Carmelina and Tsoukalas, Ioannis and Bellos, Vasilis},
  title = {A stochastic rain-on-grid framework for handling spatio-temporal rainfall uncertainty in impact-based flood nowcasting},
  journal = {International Journal of Disaster Risk Reduction},
  year = {2026},
  doi = {10.1016/j.ijdrr.2026.105998},
  url = {https://doi.org/10.1016/j.ijdrr.2026.105998}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)