Day 4

Detailed paper information

Back to list

Paper title Characterization different scattering mechanisms using SAR Polarimetry for the improvement of SAR wind retrieval processing
Authors
  1. Abdalmenem Owda Technical University of Denmark (DTU) Speaker
  2. Merete Badger Technical University of Denmark
  3. Jørgen Dall Technical University of Denmark
  4. Dalibor Cavar
Form of presentation Poster
Topics
  • A8. Ocean
    • A8.13 Remote-sensing of Ocean Winds and Stress
Abstract text Introduction
With the increased numbers of marine traffic and the man-made objects in the oceans, such as ships, oil platforms and many others, it becomes indispensable to detect these objects to benefit the maritime applications. The abundance of SAR data and its free-open availability encourage the researchers and the industry to exploit this unique remote sensing data to characterize different scatterers in oceans.
SAR wind retrieval processing depends mainly on the Bragg scattering mechanism[1], the scattering of radar pulses caused by centimeter-scale waves on top of long waves. It is possible to relate the Normalized Radar Cross Section (NRCS) values resulting from these small waves to wind speed measurements using geophysical model functions such as CMOD5[2]. Nevertheless, existence of any other type of scattering than Bragg scattering can negatively violate the accuracy of the retrieved wind speed from SAR. The wind speed accuracy matters for many applications, such as offshore wind energy applications. Therefore, quad polarized SAR data can be the key to improve the accuracy of SAR wind retrieval and create quality flags for the SAR wind maps based on different scattering mechanisms occurring in the imaged scene itself.
Different detection approaches can be used to characterize the anomalous pixels in SAR scenes. Constant False Alarm Rate (CFAR) is one of the prominent algorithms utilized to distinguish and detect ships in the ocean based on a threshold value. Nevertheless, this approach depends greatly on the background clutter distribution, subsequently; the CFAR algorithm may have severe problems at heterogeneous areas[3].

Datasets
Many satellite sensors are able to provide different polarization modes. Sentinel-1 (2014- present) collects C-band dual polarization over land worldwide as well as over priority coastal zones . PALSAR-1 (2006-2011) & 2 (2014-now), are side-looking phased array L-band SAR sensors and their Polarimetry mode (PLR) data are available. Last but not least, RADARSAT-1 (1995-2013) & 2 (2007-present) provide standard, wide and fine quad polarization mode. However, some challenges are facing these systems, among these, their technology is complex and the swath of the products is less compared with a single polarized system.

Theory
The diversity in full polarimetric datasets (POlSAR) allows us to complete characterization of the scattered objects rather than the dual and single polarized images, respectively. Each pixel in the full polarimetric scene can be represented in a scattering matrix (S), whereas its components are known as complex scattering measured as amplitudes from different combinations of V and H polarization, as follows:
S= [■(S_HH&S_HV@S_VH&S_VV )]
where S_HV is the backscattered coefficient from horizontal polarization transmission and vertical polarization reception. Other terms are similarly defined.
Scattering matrix gives information about the complete scattering process and can be directly employed to determine a deterministic or a single object scatter, but this is not the case we face near the offshore wind farms. Then, the S is random due to different scatterers that may exist in our study area. Speckle noise filtering is a crucial step in POlSAR processing to define accurately the covariance (C) and coherence matrix (T) while keeping the spatial resolution. Several general principles can be reviewed or implemented to select the best optimal principle, which fits our data. Not to mention all, one and multidimensional gaussian distribution, and the wishart distribution are used to study the distributed scatterers properties through estimation of the C and T matrix. In other words, the C and T matrix are obtained as a vectorization of the S matrix to obtain a new formulation to describe the information contained in the S matrix[4].

Methodology
This study is going to target different full and dual polarimetric datasets for offshore wind farms areas. The methodology of this study is illustrated as shown in the flowchart diagram. The methodology has the validation approach to validate the output of polarimetric decomposition outputs with conventional algorithms such CFAR, Likelihood ration test (LRT) for POlSAR ship detector, and faster Regional Convolution Neural Network(R-CNN) model. Furthermore, the non-Bragg scattering areas will be handled using a developed deep learning (DL) model to refill these areas with proper NRCS values to infer wind speed values.


Expected outcomes
Work is going on to approach the end product of the workflow: wind fields retrieved from SAR with added information about the wind speed quality. This research will definitely benefit many maritime applications and especially the offshore wind energy applications.

Acknowledgements
The PhD project belongs to the Train2Wind network. The Innovation Training Network Marie-Curie Actions: Train2Wind has received funding from the European Union Horizon 2020. Thanks to the European Space Agency for providing us with the full polarimetric datasets.

References
[1] G. R. Valenzuela, “Theories for the interaction of electromagnetic and oceanic waves - A review,” Boundary-Layer Meteorol., vol. 13, no. 1–4, pp. 61–85, 1978, doi: 10.1007/BF00913863.
[2] H. Hersbach, “Comparison of C-Band scatterometer CMOD5.N equivalent neutral winds with ECMWF,” J. Atmos. Ocean. Technol., vol. 27, no. 4, pp. 721–736, 2010, doi: 10.1175/2009JTECHO698.1.
[3] C. Liu, P. W. Vachon, R. A. English, and N. Sandirasegaram, “Ship detection using RADARSAT-2 Fine Quad Mode and simulated compact polarimetry data,” no. February, p. 74, 2010.
[4] Y. Yamaguchi, Polarimetric Synthetic Aperture Radar. 2020.