Yang et al. (2026) Multi-resolution downscaling of inundation extents using deep learning

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-06-26
• Authors: Chanyu Yang, Fiachra E. O’Loughlin
• DOI: 10.1016/j.jhydrol.2026.135932

Research Groups

• School of Civil Engineering, University College Dublin
• Dooge Centre for Water Resources Research, University College Dublin

Short Summary

This study proposes a deep learning model to downscale low-resolution flood inundation extents into high-resolution extents. The results show that the proposed method can accurately predict flood inundation and outperform traditional hydrodynamic models in terms of computational efficiency.

Objective

• Investigate the use of generative adversarial networks (GANs) for downscaling flood inundation extents.
• Evaluate the performance of GAN models with different architectures and hyperparameters.
• Compare the results with traditional hydrodynamic models, such as LISFLOOD-FP.

Study Configuration

• Spatial Scale: Urbanised area in Dublin, Ireland, covering 16.3 km2.
• Temporal Scale: Flood events from 2010 to 2011, with a temporal resolution of 15 minutes.

Methodology and Data

• Models used: Generative Adversarial Networks (GANs), LISFLOOD-FP hydrodynamic model.
• Data sources: Hydrodynamic model outputs, high-resolution Digital Terrain Model (DTM).

Main Results

• The proposed GAN model can accurately downscale low-resolution flood inundation extents into high-resolution extents.
• The results show that the GAN model outperforms traditional hydrodynamic models in terms of computational efficiency.
• The study demonstrates the potential of using GANs for real-time flood inundation forecasting.

Contributions

• This study contributes to the development of new methods for downscaling flood inundation extents.
• The proposed method can be used for real-time flood inundation forecasting, reducing the need for traditional hydrodynamic models.
• The study highlights the importance of considering the impact of model architecture and hyperparameters on performance.

Funding

• This research was funded by [insert funding agency/program/project code].

Citation

@article{Yang2026Multiresolution,
  author = {Yang, Chanyu and O’Loughlin, Fiachra E.},
  title = {Multi-resolution downscaling of inundation extents using deep learning},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135932},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135932}
}