Globally scalable glacier mapping by deep learning matches expert delineation accuracy
2024

Global Glacier Mapping Using Deep Learning

publication 10 minutes Evidence: high

Author Information

Author(s): Maslov Konstantin A., Persello Claudio, Schellenberger Thomas, Stein Alfred

Primary Institution: University of Twente, The Netherlands

Hypothesis

Can a deep learning model accurately map glaciers on a global scale using open satellite imagery?

Conclusion

The GlaViTU model achieves expert-level accuracy in glacier mapping, enhancing the reliability of glacier monitoring for climate change analysis.

Supporting Evidence

  • The model achieved an intersection over union score of over 0.85 on previously unobserved images.
  • Adding synthetic aperture radar data improved accuracy in all regions where available.
  • The study released a benchmark dataset covering 9% of glaciers worldwide.

Takeaway

Scientists created a smart computer program that can find and outline glaciers in pictures from space, helping us understand how glaciers are changing.

Methodology

The study developed a convolutional-transformer deep learning model called GlaViTU and tested it on a dataset of satellite images to map glaciers.

Potential Biases

Potential biases may arise from the training data, affecting the model's generalization to new images.

Limitations

The model struggles with debris-covered glaciers and shadowed ice, which can lead to inaccuracies.

Digital Object Identifier (DOI)

10.1038/s41467-024-54956-x

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication