Hurricane María Landslide Source Areas in Puerto Rico Identified via M3C2 Method

Puerto Rico's landslide source areas were identified using the Multiscale Model-to-Model Cloud Comparison (M3C2) method on pre- and post-Hurricane María LiDAR data. The dataset contains over 96,100 filtered landslide events across four watersheds, resulting from a master's thesis at the University of Puerto Rico at Mayagüez. Initial detection identified over 180,000 areas, which were filtered to remove those with an average depth less than 0.5 m or area smaller than 25 m².

Use Cases

• Benchmarking automated landslide detection algorithms based on the over 10,000 manually identified source areas in the Lago Caonillas watershed.
• Analyzing the spatial distribution and scale of mass wasting events triggered by Hurricane María based on the filtered dataset of over 96,100 landslides.
• Training or validating machine learning models for geohazard identification based on LiDAR-derived depth and area metrics.
• Studying watershed-specific erosion and landslide susceptibility based on data from the Lago Caonillas, Lago Dos Bocas, Lago Lucchetti, and Río Grande de Añasco basins.

Strengths

• Contains over 96,100 validated landslide source areas, providing a substantial sample for analysis.
• Data is filtered using specific geomorphic thresholds (average depth >= 0.5 m and area >= 25 m²), which likely improves reliability.
• Includes a benchmark of over 10,000 manually identified landslide source areas in the Lago Caonillas watershed for method evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown for the raw, unfiltered data, which may limit suitability assessment for some use cases.
• The dataset's specific file formats and structure are not described, requiring inspection upon access.

Provenance

Source

DesignSafe Data Depot Repository Harvested Subcollection, associated with a master's thesis from the University of Puerto Rico at Mayagüez.

Collection Method

Application of a modified Python script implementing the M3C2 automatic change detection methodology to pre- and post-Hurricane María LiDAR data.

Time Range

Likely centered around Hurricane María, which made landfall in September 2017.

Freshness

Last updated 2026-04-20 14:52:46; freshness should be verified.

Geography

Puerto Rico, specifically the Lago Caonillas, Lago Dos Bocas, Lago Lucchetti, and Río Grande de Añasco watersheds.