Floating Kelp Forest Classification Model for British Columbia Coast Using SAR

A machine learning model in .sav format for classifying floating kelp forests from spaceborne Synthetic Aperture Radar (SAR). The bundle includes sample input data for 2017 covering Malcolm Island, BC, Canada, a Jupyter notebook, and a Miniconda environment definition for testing. This algorithm resulted from a research project funded by the Canadian Space Agency and the Natural Sciences and Engineering Research Council of Canada, with details published in an associated peer-reviewed paper.

Use Cases

• Classifying floating kelp presence from SAR imagery based on the provided pre-trained model.
• Testing and adapting the kelp detection algorithm using the included Jupyter notebook and sample 2017 data.
• Researching coastal ecosystem changes by applying the model to new SAR data from the British Columbia coast.

Strengths

• Includes a complete, ready-to-use machine learning model saved in .sav format.
• Provides a reproducible testing environment with a Jupyter notebook and Miniconda definition.
• Sample input data for the year 2017 offers a concrete starting point for validation.

Limitations

• The mask for land, sandy bottom substrate, and depths under 20m is excluded due to copyright, requiring users to create their own.
• Column-level documentation is absent; field semantics must be inferred after download.
• The scale of the full training dataset and its temporal coverage beyond the 2017 sample are unknown.

Provenance

Source

Borealis Harvested Dataverse, authored by Marta Konik.

Collection Method

Developed as part of a research project using spaceborne Synthetic Aperture Radar.

Time Range

Sample data is from 2017.

Freshness

Last updated 2026-05-02 04:11:04; freshness should be verified.

Geography

Malcolm Island, British Columbia, Canada coast.