Troy et al. (2025) Can runoff modeled at coarse resolution simulate floods at finer resolutions? A case study over the Ohio River Basin

Identification

• Journal: Advances in Water Resources
• Year: 2025
• Date: 2025-10-19
• Authors: Tara J. Troy, Naresh Devineni, Carlos Lima, Upmanu Lall
• DOI: 10.1016/j.advwatres.2025.105151

Research Groups

• Department of Civil Engineering, University of Victoria, Victoria, BC, Canada
• Department of Civil Engineering, The City University of New York (City College), NY, NY, USA
• Earth and Environmental Sciences, The City University of New York (Graduate Center), NY, NY, USA
• Department of Civil and Environmental Engineering, University of Brasilia, Brasilia, DF, Brazil
• Columbia Water Center, The Earth Institute, Columbia University, NY, NY, USA
• Department of Earth and Environmental Engineering, Columbia University, NY, NY, USA
• Water Institute, Arizona State University, Tempe, AZ, USA

Short Summary

This study evaluates a computationally efficient flood modeling framework that couples the coarse-resolution Variable Infiltration Capacity (VIC) land surface model with a newly developed 1 km resolution kinematic wave routing model over the Ohio River Basin. The framework successfully reproduces flood characteristics, demonstrating that while sub-daily temporal resolution has minimal impact, coarser spatial resolutions lead to significant underestimation of flood peaks.

Objective

• To validate a modeling approach that couples the Variable Infiltration Capacity (VIC) land surface model with a high-resolution kinematic wave routing model for simulating floods in the Ohio River Basin.
• To evaluate the impact of model spatial resolution and timestep on the accuracy of flood simulations.
• Principal hypothesis: A more realistic river network with streamflow routing, even with coarse-resolution runoff data, can improve the understanding of flood-generating processes and provide a computationally efficient framework for flood risk estimation.

Study Configuration

• Spatial Scale: Ohio River Basin (525,765 km²). VIC model run at 1/8° (approximately 13 km) for baseline, with experiments at 1/4°, 1/2°, and 1° spatial resolutions. The kinematic wave routing model operates at a 1 km spatial resolution.
• Temporal Scale: January 1, 1979, through December 31, 2022 (44 years). VIC model run at an hourly timestep for baseline, with experiments at 3-hour, 6-hour, and 24-hour (daily) timesteps.

Methodology and Data

• Models used:
  • Variable Infiltration Capacity (VIC) land surface model (Liang et al., 1994)
  • Kinematic wave routing model (newly developed for this study, based on Chow et al., 1988)
• Data sources:
  • USGS streamflow gauges: Daily streamflow data from 200 gauges with drainage areas ranging from 1,002 km² to 525,765 km² and at least 20 years of data.
  • NLDAS-2 meteorological forcing dataset: Hourly meteorological data at 1/8° spatial resolution (1979–2022). Precipitation derived from CPC dataset (gauge-only) adjusted by PRISM and disaggregated using NOAA Doppler Stage II radar and CMORPH data. Other fields (air temperature, pressure, vapor pressure, wind speed, radiation) derived from NCEP North American Regional Reanalysis (NARR).
  • HydroSHEDS dataset: 1 km resolution for river network extraction and channel slope estimation.
  • USGS field measurements: Used to fit a power law for channel width as a function of drainage area.
  • US Army Corps of Engineers’ National Inventory of Dams (NID): Used to locate dams and calculate upstream storage capacity for interpreting results.

Main Results

• The modeling framework simulates the median annual maximum daily flow (AMF) with an average bias of 1.8% and the 90th percentile AMF with an average bias of 6.2% across 200 USGS gauges for 1979–2022.
• Errors are larger in flatter regions and smaller basins with dams, where the model tends to overestimate flood peaks due to the exclusion of dam management.
• The model accurately reproduces the time series of flooding, particularly in mountainous regions, but tends to be too "flashy" in flatter areas.
• Estimates of 100-year return period floods show good agreement with observations (R² = 0.85), with modeled estimates slightly lower (average underestimation of 6.5%). Discrepancies are primarily linked to errors in the variability (scale parameter) of AMF rather than the mean.
• Simulated AMF is not sensitive to sub-daily model timesteps (1, 3, or 6 hours), indicating that flood-generating mechanisms in this region are well-captured at 6-hour resolution.
• Simulated AMF is sensitive to spatial resolution, with coarser grid cells leading to increasing underestimation of AMF (1/4° by 1.2%, 1/2° by 3.8%, and 1° by 9.9% for median AMF).
• Running the VIC model at a daily timestep (24 hours) results in significantly different model behavior and larger underestimations (median AMF changes by -20%) compared to sub-daily timesteps.
• The routing model is insensitive to channel bed slope but sensitive to Manning's n and river sinuosity, suggesting potential for improvement through calibration of these parameters.

Contributions

• Demonstrates a feasible and accurate approach for computationally efficient flood modeling in large river basins by coupling a coarse-resolution land surface model (VIC) with a high-resolution (1 km) kinematic wave routing model.
• Provides a framework that overcomes the limitations of traditional coarse-resolution routing by enabling the simulation of streamflow dynamics throughout a densely represented river network, facilitating a better understanding of flood processes.
• Quantifies the trade-offs between computational efficiency and accuracy, showing that sub-daily temporal resolution changes have minimal impact on flood peaks, while coarser spatial resolutions lead to significant underestimation.
• Establishes a physically consistent framework suitable for future studies on flood-generating mechanisms, robust flood risk estimation using ensembles, and quantifying projected changes in flooding due to climate change.

Funding

• AIG Insurance Group

Citation

@article{Troy2025Can,
  author = {Troy, Tara J. and Devineni, Naresh and Lima, Carlos and Lall, Upmanu},
  title = {Can runoff modeled at coarse resolution simulate floods at finer resolutions? A case study over the Ohio River Basin},
  journal = {Advances in Water Resources},
  year = {2025},
  doi = {10.1016/j.advwatres.2025.105151},
  url = {https://doi.org/10.1016/j.advwatres.2025.105151}
}