Global Flood Susceptibility Map (GFSM v1): A high resolution (30m) flood susceptibility dataset derived from multi-source geospatial data

The Global Flood Susceptibility Map (GFSM v1) is the first globally harmonized, high-resolution (30 m) flood susceptibility dataset derived using a machine learning framework. It identifies areas inherently prone to flooding based on landscape and environmental characteristics, independent of specific flood event frequency or magnitude.

This dataset was produced by training gradient-boosted tree model (XGBoost) on approximately 30.45 million samples extracted from ~17,000 grid tiles worldwide (~10,000 GBs of data, ~1.7 trillion pixels at 30m resolution). The models utilized global flood inundation data (Aqueduct Flood Hazard Maps) as training labels and incorporated nine flood conditioning factors:

• Elevation

• Slope

• Aspect

• Topographic Wetness Index (TWI)

• Height Above Nearest Drainage (HAND)

• Vegetation Index (NDVI)

• Distance to water bodies

• Distance to roads

• Rainfall frequency

Dataset Specifications

• Spatial Resolution: 30 meters (approx. 0.000269 degrees)

• Geographic Coverage: Global (approx. 80°N to 60°S)

• Coordinate Reference System (CRS): World Mercator (EPSG:3395)

• File Format: GeoTIFF, tiled

• Model Performance: Median ROC-AUC ~0.95 and F1-score ~0.90 across 192 regional units

Classification Scheme

The dataset provides susceptibility classes derived from the model's continuous probability output using equal-interval slicing. The pixel values correspond to the following categories:

File Structure & Access

Due to the large data volume, the dataset is organized into regional compressed archives. Users should follow this workflow:

1. Download the Tile Index: Download GFSM_Tile_Index.zip first. This contains a GeoPackage (.gpkg) index.

2. Identify Your Region: Open the index in a GIS (QGIS or ArcGIS) to identify the specific Tile ID and the corresponding Regional Zip file for your area of interest.

3. Download Regional Data: Locate and download the specific archive (e.g., GFSM_N20W020.zip). These archives contain the GeoTIFF tiles and associated metadata JSON files.

Usage Notes

• Application: GFSM v1 represents a baseline of flood susceptibility under present-day conditions. It is intended for exposure assessment, risk screening, and comparative spatial analysis.

• Limitations: The map reflects the natural propensity of the landscape to flood. It does not account for man-made dynamic flood defenses (e.g., levees) or future climate change projections beyond the training data baseline.

Funding & Acknowledgments

This research was supported by Hong Kong Baptist University (HKBU) and utilized the Shaheen supercomputing resources at King Abdullah University of Science and Technology (KAUST).

Web Application:

The dataset is already available in Google Earth Engine as image collection. For details please check the Project GitHub repository at GFSM GitHub.

Moreover, for quick visualization, the data is made available in the form of GEE web application, which can be accessed at GFSM Web App.