Automated Mapping of Supraglacial Stream Networks on the Greenland Ice Sheet Using Dual U-Net Convolutional Neural Networks
Abstract
Detection and monitoring of supraglacial meltwater drainage patterns is critical for accurately predicting ice mass loss from the Greenland Ice Sheet. However, manual digitizing is time consuming and imperfect, and traditional automated classifications contain physically unrealistic discontinuities. We present a novel dual U-Net architecture for automated mapping of supraglacial meltwater channels on the Greenland Ice Sheet using satellite imagery that decreases delineation time for an image from hours to
Fields of Study & Subjects
Cryospheric studies and observations
Polar Research and Ecology
Arctic and Antarctic ice dynamics
Publication details
- Journal / source
- IEEE Geoscience and Remote Sensing Letters
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Volume
- 23
- Pages
- 1–5
- Publication date
- 1 January 2026