Awasthi et al. (2026) Conditional simulation for max-stable processes (CoSim-MSP): A framework for flood prediction at ungauged basins

Identification

• Journal: Advances in Water Resources
• Year: 2026
• Date: 2026-08-03
• Authors: Chandramauli Awasthi, A. Sankarasubramanian
• DOI: 10.1016/j.advwatres.2026.105449

Research Groups

• Department of Civil, Construction, and Environmental Engineering, North Carolina State University, USA

Short Summary

This study proposes a new framework called Conditional Simulation of Max-Stable Process (CoSim-MSP) to predict floods at ungauged basins based on observed extremes at neighboring gauges. The framework utilizes max-stable models that can produce unconditional simulations for a group of basins in a region and resamples conditional simulations from them, preserving spatial dependence structure.

Objective

• Develop a computationally efficient framework for conditional flood simulation using max-stable processes to predict floods at ungauged basins based on observed extremes at neighboring gauges.

Study Configuration

• Spatial Scale: Regional scale (HUC04 regions in the United States)
• Temporal Scale: 50-year period (1971-2020)

Methodology and Data

• Models used: Max-stable processes, specifically Brown-Resnick model, Power-exponential model, and Cauchy model.
• Data sources: Global River Data Centre (GRDC) dataset, USGS stream gauge data.

Main Results

• The CoSim-MSP framework demonstrates strong predictive skill for 23 out of 35 basins with R² > 0.50.
• The framework's performance is directly proportional to the strength of spatial dependence among observed AMFs.
• The poor performance of the model for certain basins can be attributed to factors controlling the spatial coherence of floods, such as climatological and basin's physical characteristics-based heterogeneity.

Contributions

• This study addresses the challenge of conditional flood simulation using max-stable processes and proposes a new framework that preserves spatial dependence structure.
• The CoSim-MSP framework provides median flood prediction estimate along with prediction uncertainty, making it more reliable than conventional approaches.

Funding

• Not specified in the paper.

Citation

@article{Awasthi2026Conditional,
  author = {Awasthi, Chandramauli and Sankarasubramanian, A.},
  title = {Conditional simulation for max-stable processes (CoSim-MSP): A framework for flood prediction at ungauged basins},
  journal = {Advances in Water Resources},
  year = {2026},
  doi = {10.1016/j.advwatres.2026.105449},
  url = {https://doi.org/10.1016/j.advwatres.2026.105449}
}