Sea-Ice Floe Segmentation Products from MEDEA Imagery, 1999-2014

78 MEDEA satellite images from the USGS Global Fiducials Library, processed to identify individual sea-ice floes in the Canada Basin. The segmentation products were developed by Alexis Denton at Yale University for a 2022 research paper funded by the Office of Naval Research. This multidecadal dataset provides identified floe characteristics like size and shape for Arctic science.

Use Cases

• Analyzing sea-ice floe size distribution over time based on the 15-year time series
• Training computer vision models for ice floe segmentation based on the processed image products
• Studying changes in Arctic sea ice morphology based on geospatial imagery from the Canada Basin

Strengths

• Covers a 15-year period from 1999 to 2014, enabling multidecadal analysis
• Derived from 78 high-resolution MEDEA satellite images
• Includes the segmentation algorithm code, hosted on GitHub with a Zenodo DOI

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count and file size are unknown, which may limit suitability assessment
• Data may reflect geographic bias inherent to the Canada Basin region

Provenance

Source

United States Geological Survey (USGS) Global Fiducials Library

Collection Method

Algorithmic segmentation of satellite imagery

Time Range

1999 through 2014

Geography

Canada Basin, Arctic