Detection and classification of sea ice from spaceborne multi-frequency synthetic aperture radar imagery and radar altimetry
Doctoral thesis, 2020
The work presented in this thesis focuses on the one hand on synergies of multi-frequency spaceborne synthetic aperture radar (SAR) imagery for sea ice classification. On the other hand, the fusion of radar altimetry observations with near-coincidental SAR imagery is investigated for its potential to improve 3-dimensional sea ice information retrieval.
Investigations of ice/water classification of C- and L-band SAR imagery with a feed-forward neural network demonstrated the capabilities of both frequencies to outline the sea ice edge with good accuracy. Classification results also indicate that a combination of both frequencies can improve the identification of thin ice areas within the ice pack compared to C-band alone. Incidence angle normalisation has proven to increase class separability of different ice types. Analysis of incidence angle dependence between 19-47° at co- and cross-polarisation from Sentinel-1 C-band images closed a gap in existing slope estimates at cross-polarisation for multiyear sea ice and confirms values obtained in other regions of the Arctic or with different sensors. Furthermore, it demonstrated that insufficient noise correction of the first subswath at cross-polarisation increased the slope estimates by 0.01 dB/1° for multiyear ice. The incidence angle dependence of the Sentinel-1 noise floor affected smoother first-year sea ice and made the first subswath unusable for reliable incidence angle estimates in those cases.
Radar altimetry can complete the 2-dimensional sea ice picture with thickness information. By comparison of SAR imagery with altimeter waveforms from CryoSat-2, it is demonstrated that waveforms respond well to changes of the sea ice surface in the order of a few hundred metres to a few kilometres. Freeboard estimates do however not always correspond to these changes especially when mixtures of different ice types are found within the footprint. Homogeneous ice floes of about 10 km are necessary for robust averaged freeboard estimates.
The results demonstrate that multi-frequency and multi-sensor approaches open up for future improvements of sea ice retrievals from radar remote sensing techniques, but access to in-situ data for training and validation will be critical.
sea ice classification
sea ice concentration
Chalmers, Space, Earth and Environment, Microwave and Optical Remote Sensing
Sea ice concentration estimation from Sentinel-1 Synthetic Aperture Radar images over the Fram Strait
36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016; Beijing; China; 10-15 July 2016,; Vol. 2016-November(2016)p. Art no 7731001, Pages 7675-7677
Paper in proceedings
Comparison of ice/water classification in Fram Strait from C- A nd L-band SAR imagery
Annals of Glaciology,; Vol. 59(2018)p. 112-123
Aldenhoff, W., Eriksson, L. E. B., Ye, Y. and Heuzé, C. First-year and Multiyear Sea Ice Incidence Angle Normalization of Dual-polarized Sentinel-1 SAR Images in the Beaufort Sea
Comparison of Sentinel-1 Sar and Sentinel-3 Altimetry Data for Sea Ice Type Discrimination
International Geoscience and Remote Sensing Symposium (IGARSS),; (2019)p. 4238-4240
Paper in proceedings
Sensitivity of radar altimeterwaveform to changes in sea ice type at resolution of synthetic aperture radar
Remote Sensing,; Vol. 11(2019)
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4729
Chalmers University of Technology
Opponent: Dr. Marko Mäkynen, Finnish Meteorological Institute, Helsinki, Finland