1. Weimin Huang, Yali Wang Memorial University, Canada
A Spectra-Analysis-Based Method for Wind Measurements Using X-Band Marine Radar
Weimin Huang, Yali Wang
Memorial University, Canada

2. Outline 1. Introduction 2. Spectra-Analysis-Based Method 3. Results
2.1 Wind direction
2.2 Wind speed
3. Results
4. Conclusions

3. Why Remote Sensing Sea Surface Wind?
1 Introduction
Why Remote Sensing Sea Surface Wind?
Study energy exchange processes between the atmosphere and the ocean surface
Enhance marine navigation safety
Finance wind farms project
Investigate the influence on the ecosystem

4. Existing Wind Algorithms
1. Introduction
Existing Wind Algorithms
Streaks-based (2003, Dankert and Horstmann)
Curve-fitting-based (2012, Lund et al)
Intensity-level-selection-based (2013, Vicen-Bueno et al)
Spectra-analysis-based (2016, Wang and Huang)

5. 2.1 Wind Direction Principle Fitting function (Lund et al)
HH-polarized X-band marine radar operating at grazing incidence, radar backscatter has a single peak in the upwind direction.
Fitting function (Lund et al)
𝜎 𝜃 :average intensity
𝜃: antenna look direction
𝑎 2 : upwind direction
A radar image
Fitting result

6. 2.1 Wind Direction Problems under low-wind-speed rain conditions
A rain contaminated image
(wind speed: 4.2 m/s)
(Wind direction: 266°)
Curve fitting results (Retrieved wind direction: 87°)

7. 2.1 Wind direction Wavenumber domain based method Steps:
1. Fourier transform
𝑁: pixel number in one direction
𝐼 𝜃 (𝑛): 𝑛th intensity in direction 𝜃
𝐸 𝜃 (𝑘): spectral value at wavenumber 𝑘
Wavenumber spectra in one direction

8. Wavenumber domain based method
2.1 Wind Direction
Wavenumber domain based method
Steps (cont.):
2. Normalization
3. Obtain 𝑆 𝜃
𝑆 𝜃 : reflect intensity variation
Wavenumber spectra
k∈ [0.01, 0.2]

9. 2.1 Wind Direction Wavenumber domain based method Steps (cont.):
4. Least-square fitting
𝑎 2 : upwind direction
Results using new method
(Retrieved wind direction: 254°)

10. High-wind-speed rain cases
2.1 Wind Direction
Wavenumber domain based method
High-wind-speed rain cases
Rain-free cases Sθ
No single peak
Single peak in upwind direction
Eθ(0)
Thus, for rain-free or rain data under high wind speed:
a2: upwind direction

11. Curve-fitting-based method
2.2 Wind Speed
Curve-fitting-based method
Average intensity
σave
Wind speed Ws
Can’t be applied to rain-contaminated data
Unstable

12. 2.2 Wind Speed Wavenumber domain based method
Wavenumber spectra in one direction
Zero and nonzero components

13. Wavenumber domain based method
2.2 Wind Speed
Wavenumber domain based method
Fourier transform
each azimuth data to
wavenumber domain
Integrate all
spectral
components
where Δr: range resolution
Curve fitting S
and wind speed
Blockdiagram

14. Radar system specification
3.1 Results
Radar data information
Radar system specification
Collecting date
Nov-Dec, 2008
Polarization HH
Radar location
42.5º N, 62.08º W (off Halifax)
Antenna rotation speed
28 rpm
Range resolution
7.5 m

15. 3. Results Wind Direction (a) (b)
(a) Retrieved wind direction results; (b) measured rain rates

16. 3. Results Wind Direction RMSE Curve fitting method New method
RMSE for low-wind-speed rain-contaminated data
46.7°
21.6°
RMSE for data in other cases
14.9°

17. Wavenumber domain based method
3. Results
Wind speed Model
Curve fitting method
Wavenumber domain based method

18. 3. Results Wind Speed (a) (b)
(a) Retrieved wind speed results; (b) measured rain rates

19. Wavenumber domain based method
3. Results
Wind Speed RMSE
Curve fitting method
Wavenumber domain based method
Training data
& , rain-free, wind speed > 2 m/s
RMSE for all rain-contaminated data
7.5 m/s
1.6 m/s
RMSE for all rain-free data
1.5 m/s

20. 4. Conclusions A wavenumber domain based method (NM1)
A gamma correction incorporated method (NM2)
Model: logarithmic function
Rain cases: accuracy improved by 5.9 m/s (NM1) & 5.4 m/s (NM2)
Rain-free cases: accuracy almost the same with CF method
A wavenumber domain based method
Low-wind-speed rain cases: accuracy improved by 25.1°
High-wind-speed rain & rain-free cases: accuracy maintain the same with CF method

21. References A wavenumber domain based method (NM1)
A gamma correction incorporated method (NM2)
Model: logarithmic function
Rain cases: accuracy improved by 5.9 m/s (NM1) & 5.4 m/s (NM2)
Rain-free cases: accuracy almost the same with CF method
[1] B. Lund, H. C. Graber, and R. Romeiser, “Wind retrieval from shipborne nautical X-band radar data,” IEEE Trans. Geosci. Remote Sens., vol. 50, no. 10, pp , Oct
[2] H. Dankert and J. Horstmann, “A marine radar wind sensor,” J. Atmos. Ocean. Technol., vol. 24, no. 9, pp , Sep
[3] R. Vicen-Bueno, J. Horstmann, E. Terril, T. De Paolo, and J. Dannenberg, “Real-time ocean wind vector retrieval from marine radar image sequences acquired at grazing angles,” J. Atmos. Ocean. Technol., vol. 30, no. 1, pp , Jan
[4] M. Tsimplis and S. A. Thorpe, “Wave damping by rain,” Nature, vol. 342, pp , 1989.
[5] R. H. Wanninkhof, L. F. Bliven, “Relationship between gas exchange, wind speed, and radar backscatter in a large wind- wave tank,” J. Geophys. Res., vol. 96, no. C2, pp , 1984.
[6] W. Huang and E. W. Gill, “Ocean remote sensing using X-band shipborne nautical radar - applications in eastern Canada,” in Coastal Ocean Observing Systems: Advances and Syntheses, 1st ed., Canada: Elsevier Press, 2015, Ch. 15, pp
[7] C. A. Poynton, “Gamma,” in Digital Video and HDTV: Algorithms and Inter- faces, Amsterdam, Netherlands: Morgan Kaufmann Publishers, 2003, ch. 23, pp

22. Acknowledgement
1. DRDC (Halifax) for providing X-band radar filed data.
2. NSERC Discovery grants support.
3. Memorial University SBM grant support.

23. Thank you!
Any questions?