School of Marine Sciences, Nanjing University of Information Science & Technology, Nanjing 210044, China
2.
East Sea Information Center of State Oceanic Administration, Shanghai 200136, China
Funds:
The National Key Research and Development Program of China under contract No. 2021YFC2803301; the National Natural Science Foundation of China under contract No. 41977302; the National Natural Science Youth Foundation of China under contract No. 41506199, the Natural Science Youth Foundation of Jiangsu Province under contrant No. BK20150905 and Science and Technology Project of China Huaneng Group Co., Ltd. under contract NO. HNKJ20-H66
Sea ice surface roughness (SIR) affects the energy transfer between the atmosphere and the ocean, and it is also an important indicator for sea ice characteristics. To obtain a small-scale SIR with high spatial resolution, a novel method is proposed to retrieve SIR from Sentinel-1 synthetic aperture radar (SAR) images, utilizing an ensemble learning method. Firstly, the two-dimensional continuous wavelet transform is applied to obtain the spatial information of sea ice, including the scale and direction of ice patterns. Secondly, a model is developed using the Adaboost Regression model to establish a relationship among SIR, radar backscatter and the spatial information of sea ice. The proposed method is validated by using the SIR retrieved from SAR images and comparing it to the measurements obtained by the Airborne Topographic Mapper (ATM) in the summer Beaufort Sea. The determination of coefficient, mean absolute error, root-mean-square error and mean absolute percentage error of the testing data are 0.91, 1.71 cm, 2.82 cm and 36.37%, respectively, which are reasonable. Moreover, K-fold cross-validation and learning curves are analyzed, which also demonstrate the method's applicability in retrieving SIR from SAR images.
Figure 1. Map of the sea ice age from National Snow and Ice Data Center (NSIDC) (Tschudi et al., 2019) on July 15, 2016. Location of the study area is denoted as a white rectangle.
Figure 2. Preprocessed Sentinel-1 SAR image in HH-polarization sensed on July 13, 2016.
Figure 3. Sentinel-1 SAR HH-polarization backscatter coefficient (σ0) image under different incidence angles. a. Small incidence angle. b. Middle incidence angle. c. Large incidence angle.
Figure 4. Spatial distribution of orthometric height and SIR in OIB ATM data. a. ATM orthometric height. b. Surfuce ice roughness.
Figure 5. Workflow of the study.
Figure 6. Relationship among SIR, $ \sigma_{0} $ and $ \theta $.
Figure 7. Scatter plots of the SIR measured by ATM and that estimated from SAR pixels.
Figure 8. Scatter plots of the SIR measured by ATM and that estimated from SAR with 2-D CWT.
Figure 9. Results of 2-D Cauchy CWT. a. Region with ice floes (Area 4 in Fig. 2). b. Gradient of Fig. 9a. (c) Peak scales. (d) Peak angles.
Figure 10. Learning curves for different types of input data. Input data without 2-D CWT (a) and input data with 2-D CWT (b). The score chosen is the coefficient of determination R2. The highlighted regions demonstrate 1 standard deviation error from the mean score of each fold.
Figure 11. The spatial distribution of samples in Center Arctic. Location of the independent test area is denoted as a blue rectangle.
Figure 12. Scatter plots of the SIR in independent test region measured by ATM and that estimated from SAR with 2-D CWT.