Nikrou et al. (2026) Intercomparison of flood inundation models across land use types and hydrological flood stages

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-03-29
• Authors: Parvaneh Nikrou, Sadra Seyvani, Sagy Cohen, Sudershan Gangrade, Joseph L. Gutenson, Dan Tian, Anupal Baruah, Michael L. Follum, Shih-Chieh Kao
• DOI: 10.1016/j.jhydrol.2026.135410

Research Groups

• Department of Geography and the Environment, University of Alabama, Tuscaloosa, AL, United States
• Oak Ridge National Laboratory, Oak Ridge, TN, United States
• Follum Hydrologic Solutions, LLC, Casper, WY, United States

Short Summary

This study provides a context-stratified intercomparison of five flood inundation models across multiple hydrograph phases, land-use types, and benchmark datasets, revealing that model performance rankings systematically shift depending on the evaluation context. The findings offer guidance for model selection and the design of future intercomparison studies.

Objective

• To provide a context-stratified intercomparison of five Flood Inundation Mapping (FIM) approaches across multiple hydrograph phases, land-use/land-cover (LULC) classes, and benchmark types to offer transferable guidance for model selection and for designing large-scale, benchmark-aware FIM intercomparison studies.

Study Configuration

• Spatial Scale: Neuse River Basin, North Carolina, USA.
• Temporal Scale: Hurricane Matthew flood (2016), assessed across two rising and two falling hydrograph limbs.

Methodology and Data

• Models used: TRITON, LISFLOOD-FP, HEC-RAS 2D, ARC-Curve2Flood, OWP HAND-FIM.
• Data sources: High-resolution remote sensing-derived flood inundation maps, hand-labeled points, and building footprints (used as validation benchmarks).

Main Results

• Model rankings for flood inundation mapping performance systematically shift across different contexts, including hydrograph phase, land-use/land-cover (LULC) type, and benchmark dataset.
• LISFLOOD-FP generally ranked highest in three of four flood phases and performed best in vegetated areas.
• TRITON led during one rising limb phase and ranked highest against hand-labeled points.
• HEC-RAS 2D showed improved relative performance in agricultural and urban areas.
• The choice of benchmark significantly influenced conclusions, with LISFLOOD-FP excelling in flooded-building detection during the late falling limb.
• Representative wall-clock runtimes were reported for each workflow, providing context for operational feasibility.

Contributions

• Provides a comprehensive, context-stratified intercomparison of five diverse FIM models, extending beyond typical single-snapshot assessments to include multiple hydrograph phases, LULC classes, and benchmark types.
• Offers transferable, context-dependent guidance for FIM model selection, which is crucial for operational decision-making and resource allocation.
• Highlights the systematic variability in model performance rankings, emphasizing the importance of context-aware evaluation in FIM intercomparison studies.
• Informs the design of more robust and benchmark-aware large-scale FIM intercomparison studies.

Funding

Not specified in the provided text.

Citation

@article{Nikrou2026Intercomparison,
  author = {Nikrou, Parvaneh and Seyvani, Sadra and Cohen, Sagy and Gangrade, Sudershan and Gutenson, Joseph L. and Tian, Dan and Baruah, Anupal and Follum, Michael L. and Kao, Shih-Chieh},
  title = {Intercomparison of flood inundation models across land use types and hydrological flood stages},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135410},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135410}
}