Lab (2025) Inundation2Depth Dataset

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Open MIND
• Year: 2025
• Date: 2025-10-09
• Authors: Geospatial and Remote Sensing Lab
• DOI: 10.5281/zenodo.17308287

Research Groups

The specific research groups, labs, or departments involved are not explicitly mentioned in the provided text.

Short Summary

This paper introduces Inundation2Depth, a novel dataset designed to overcome the scarcity of georeferenced flood depth data for deep learning applications in urban areas, providing paired inundation extent-depth labels derived from aerial imagery and LiDAR-based Digital Terrain Models.

Objective

• To address the critical lack of large quantity, georeferenced, well-labelled flood depth datasets, which constrains the development of deep learning models for urban flood severity assessment and operational mapping beyond inundation extent.

Study Configuration

• Spatial Scale: Urban settings, with data provided as scene-level raster and 256×256 tiles.
• Temporal Scale: Static dataset representing flood events at specific times of aerial imagery and LiDAR acquisition.

Methodology and Data

• Models used: HEC-RAS Rain-on-Grid tool (for dataset validation).
• Data sources: Aerial imagery, LiDAR-based Digital Terrain Models (DTMs), multi-sensor remote sensing (for complementary layers such as DTM-derived terrain features like slope and curvature, and land surface characteristics like impervious surfaces and vegetation).

Main Results

• The "Inundation2Depth" dataset was created, pairing inundation extent-depth labels derived from aerial imagery and LiDAR-based DTMs under hydrostatic assumptions.
• The dataset includes complementary layers from multi-sensor remote sensing.
• It comprises a total of 5,925 overlapping 256×256 tiles, provided in Raw and Normalized versions with consistent naming.
• The generated dataset was validated through a hydrodynamic modeling approach using the HEC-RAS Rain-on-Grid tool.

Contributions

• Lowers the data barrier for GeoAI research focused on urban flood severity.
• Enables benchmarking and training of depth-regression models specifically for urban environments.
• Facilitates ablation studies to assess the utility of various features in flood depth prediction.
• Supports the development of reproducible workflows for urban hazard assessment.
• Promotes comparability across different methods and study areas in flood modeling.

Funding

Not specified in the provided text.

Citation

@article{Lab2025Inundation2Depth,
  author = {Lab, Geospatial and Remote Sensing},
  title = {Inundation2Depth Dataset},
  journal = {Open MIND},
  year = {2025},
  doi = {10.5281/zenodo.17308287},
  url = {https://doi.org/10.5281/zenodo.17308287}
}