Articles | Volume 18, issue 5
https://doi.org/10.5194/tc-18-2207-2024
https://doi.org/10.5194/tc-18-2207-2024
Research article
 | 
03 May 2024
Research article |  | 03 May 2024

SAR deep learning sea ice retrieval trained with airborne laser scanner measurements from the MOSAiC expedition

Karl Kortum, Suman Singha, Gunnar Spreen, Nils Hutter, Arttu Jutila, and Christian Haas

Related authors

Sea Ice Freeboard Extrapolation from ICESat-2 to Sentinel-1
Karl Kortum, Suman Singha, and Gunnar Spreen
EGUsphere, https://doi.org/10.5194/egusphere-2024-3351,https://doi.org/10.5194/egusphere-2024-3351, 2024
Short summary

Cited articles

Boulze, H., Korosov, A., and Brajard, J.: Classification of Sea Ice Types in Sentinel-1 SAR Data Using Convolutional Neural Networks, Remote Sens., 12, 2165, https://doi.org/10.3390/rs12132165, 2020. a, b
Doulgeris, A. P.: An Automatic 𝒰-Distribution and Markov Random Field Segmentation Algorithm for PalSAR Images, IEEE T. Geosci. Remote, 53, 1819–1827, https://doi.org/10.1109/TGRS.2014.2349575, 2015. a
Fily, M. and Rothrock, D. A.: Extracting Sea Ice Data from Satellite SAR Imagery, IEEE T. Geosci. Remote, GE-24, 849–854, https://doi.org/10.1109/TGRS.1986.289699, 1986. a
Fritz, T., Mittermayer, J., Schaettler, B., Buckreuss, S., Werninghaus, R., and Balzer, W.: Level 1b Product Format Specification, DLR: TerraSAR-X Ground Segment, https://www.intelligence-airbusds.com/files/pmedia/public/r460_9_030201_level-1b-product-format-specification_1.3.pdf (last access: November 2022), 2007. a
Geldsetzer, T. and Yackel, J. J.: Sea ice type and open water discrimination using dual co-polarized C-band SAR, Can. J. Remote Sens., 35, 73–84, https://doi.org/10.5589/m08-075, 2009. a
Download
Short summary
A dataset of 20 radar satellite acquisitions and near-simultaneous helicopter-based surveys of the ice topography during the MOSAiC expedition is constructed and used to train a variety of deep learning algorithms. The results give realistic insights into the accuracy of retrieval of measured ice classes using modern deep learning models. The models able to learn from the spatial distribution of the measured sea ice classes are shown to have a clear advantage over those that cannot.
Share