During the yearlong MOSAIC expedition (2019-2020) a significant number of synthetic aperture radar (SAR) images were collected from different sensors and in different modes. Here, we investigate the change in polarimetric features over sea ice from the freeze up to the advanced melt season using fully polarimetric L-band images from the ALOS-2 PALSAR-2 and C-band images from the RADARSAT-2 satellite SAR sensors. The sea ice is separated into four different sea ice types: (1) lead, (2) young (YI), (3) level (second-year ice (SYI) and first year ice (FYI) and (4) deformed sea ice.
Data and Method
R/V Polarstern drifted with two different floes and here we focus on the first drift that took place between 1 October 2019 and 31 July 2020. Areas of all four different ice types are observed in the vicinity of R/V Polarstern. These areas are included whenever possible in the yearlong time series of sea ice types. Though to densify the time series images not containing the ship are also included.
The SAR images were analyzed for seasonal changes in backscatter intensity values, and the scattering mechanisms were employed to further investigate separability between the different ice types. In particular there was a focus on the separability between the high backscatter older sea ice, typically SYI, and YI and FYI. Both sets of L- and C-band images were radiometrically calibrated using the included meta-data information and a 9 × 9-pixel median filter is applied to the data to reduce the noise effect before extraction of the polarimetric features. The normalized radar backscatter information for the HH, HV and VV channels were extracted together with the polarization difference (PD, VV-HH on a linear scale), the co-polarized and cross-polarized ratios (VV/HH and HH/HV) and the circular correlation coefficient. The images were incidence angle corrected to 35◦ using the method and slope values outlined in Mahmud et al. (2018), and the different sea ice types were identified using manually drawn regions of interests (ROIs). Data from helicopter borne instruments as well as in-situ data was used to evaluate the different sea ice types. The results are also compared to images collected during the Norwegian Young Ice (N-ICE) campaign in January-June 2015 (Johansson et al, 2017, 2018).
Results and discussion
Analyzing the backscatter values several observations can be made: (i) as expected we observe that there is a larger difference in the co-polarization channels between smooth and deformed ice in L-band compared to C-band during the freezing season, though (ii) this separation is significantly reduced during the early melt season. Moreover, we observe (iii) larger differences between young ice and deformed ice backscatter values in the L-band data compared to the C-band data, and (iv) linear kinematic features (LKF) are easier to detect in the L-band images. Throughout the year the HV-backscatter values show larger differences between level and deformed sea ice in L-band than C-band. The L-band data variability is significantly smaller for the level sea ice compared to the deformed sea ice, and this variability was also smaller than that observed for the overlapping C-band data. Thus L-band data could be more suitable to reliable separate deformed from level sea ice areas, as well as investigating the LKFs.
Within the L-band images a noticeable shift towards higher backscatter values in early melt season compared to the freezing season for all polarimetric channels is observed. Though no such strong trend is found in the C-band data. The change in backscatter values is first noticeable in the C-band images and later followed by a change in the L-band images, probably caused by their different penetration depth and volume scattering sensitivities. This change also results in a smaller backscatter variability for all polarimetric channels.
PD show a seasonal dependency for the smooth and deformed sea ice within the L-band data. For the L-band data were the PD variability for all ice classes reasonably small for the freezing season, with a significant shift towards larger variability during the early melt season, though during this period the mean PD values remained similar. However, once the temperatures reached above 0°C both the variability and the mean values increased significantly. For the C-band data no such trend is observed. However, for C-band the absolute PD values show significantly higher mean values for the thinner sea ice areas regardless of if these areas are low or high backscatter, and these areas also have low standard deviations. Compared to the high backscatter areas offered by SYI there is a significant difference were the PD values have a high standard deviation but a low mean value. Using PD, we can therefore separate out the young ice types from the surrounding sea ice and the SYI types. PD is also suitable for separation between the level and deformed sea ice areas during the freeze-up, as the variability is much higher for the deformed sea ice areas than for the level ice areas. To confirm the roughness level from the different ice types the circular correlation coefficient (CCC) is calculated, compared to airborne laser scanner (ALS) data and show good separability between the deformed and level sea ice types. However, CCC is sensitive to the signal-to-noise levels and care must be taken when analyzing the results. PD on the other hand has a small incidence angle dependency and a low sensitivity to the signal-to-noise ratio (SNR).
We observe that fully polarimetric C- and L-band data are complementary to one another and that through their slightly different dependencies on season and sea ice types, a combination of the two frequencies can aid improved sea ice classification. The availability of a high spatial and temporal resolution dataset combined with in-situ information offered during the MOSAiC expedition ensures that seasonal changes can be fully explored.
Johansson A.M., C. Brekke, G. Spreen, J. King, 2018, X-, C-, and L-band SAR signatures of newly formed sea ice in Arctic leads during winter and spring, Remote Sensing of Environment, 204: 162-180
Johansson A.M, King J.A., Doulgeris A.P., Gerland S., Singha S., Spreen G., Busche T, 2017, Combined observations of Arctic sea ice with near-coincident co-located X-band, C-band and L-band SAR satellite remote sensing and helicopter-borne measurements, JGR-Oceans, 122: 669-691
M. S. Mahmud, T. Geldsetzer, S. E. L. Howell, J. J. Yackel, V. Nandan and R. K. Scharien, 2018, Incidence Angle Dependence of HH-Polarized C- and L-Band Wintertime Backscatter Over Arctic Sea Ice, IEEE Transactions on Geoscience and Remote Sensing, 56(11): 6686-6698