OC 3570 Project Brief

1
OC 3570 Project Brief

• Comparison of Wave Information Extracted by
• HF-Radar and Waverider Buoy
• References
• Barrick Lipa 1986 2nd-Order Shallow-Water
  Hydodynamic Coupling Coefficient in
  Interpretation of HF Radar Sea Echo
• Lipa Nyden Directional Wave Information from
  the SeaSonde

2
Scope of Presentation

• Introduction
• Methodology of Wave Information Extraction by
  Waverider buoy HF Radar
• Overview of Experiment Area
• Data Collection
• Key HF Radar Parameters
• Data Available from Cruise
• Data Comparison
• Time Series
• Waves Spectra
• Discussion
• Measure of Performance Summary
• SNR Analysis
• Conclusion Recommendations

3
? Introduction Wave Extraction Methodology

• Waverider Buoy

• Accelerometer (Heart of Buoy)
• horizontally stabilised platform suspended in
  fluid filled sphere to record heave (up down)
  motions
• Directional WaveRider Buoys
• Additional two fixed orthogonal accelerometers to
  measure accelerations in the north-south and
  east-west axes. This information is combined with
  the heave to resolve wave directions.

4
? Introduction Wave Extraction Methodology

• Waverider Buoy
• Provides best in-situ/ground truth straight
  forward extraction of wave information
• Provides
• Significant Wave Height
• Peak Wave Period
• Wave Direction (magnetic ? true)

5
? Introduction Wave Extraction Methodology

• HF Radar
• Hourly averages of spectral returns from 1 range
  cell (shaded) between RH LH edges
• Performs wave extraction algorithm analysis
• Wave direction
• Significant wave height
• Wave period

6
? Introduction Wave Extraction Methodology

• HF Radar
• Uses radar spectral theory
• Separate 1st order from 2nd order

Doppler Backscatter Spectrum
7
? Introduction Wave Extraction Methodology

• HF Radar
• Uses radar spectral theory Barrick, 1972a
• Separate 1st order from 2nd order
• Extract ocean wave directional spectrum at Bragg
  wave no. (i.e. 0.55 m-1 _at_ 13 MHz) from 1st order
  peaks
• Compute Fourier angular coefficients of
  broad-beam returns over the range cell
• As Bragg wave no. is small, waves assumed to
  follow the wind ? hence direction estimated from
  directional distribution defined by Fourier
  coefficients
• Derive total ocean wave spectrum
• Compute 2nd order Fourier coefficients to give
  parameters of total ocean wave spectrum
• Interpret 2nd order spectrum by either Integral
  Inversion or Model Fitting (Pierson-Moskowitz)

8
? Introduction Wave Extraction Methodology

• Assumptions
• Ocean wave spectrum is homogenous over radar
  range cell ? close-in range cell used
• Consider only deep water conditions ignore wave
  refraction
• Wind waves follow direction of winds

• Limitations
• Saturation limit for significant waveheight, for
  13MHz, hsat 7.4m
• Separate 2nd order from 1st order and noise floor
• Current velocity lt 4m/s
• User-defined software parameters
• Performs waves-averaging within range cell

Well defined 1st and 2nd order spectra
Poorly defined 1st and 2nd order spectra
9
? Introduction Overview of Experiment Area
Waverider buoy
Pt Pinos

• Southern part of Monterey Bay, with varying
  bathymetry
• Convergence of waves towards Pt Pinos ?
  contravene HF radars assumptions
• Wave height increases but at different magnitudes
  due to refraction ? cause challenges in spectra
  averaging process
• Attempt to establish waves extraction technique
  which gives more accurate HF-derived waves
  measurements

10
? ? Data Collection

• Key HF Radar parameters
• Centre Operating Freq 13.4 MHz
• Bragg freg 0.37 Hz
• Bragg Wavelength 11.2 m
• Left/Right Hand edge 225 to 075 deg

Waverider buoy
3
2
1
Pt Pinos

• Distance of HF Radar to Buoy 6.71 km
• Buoy within 2nd range cells
• 3 nearest range cells (depth 3.034km) for
  comparison purposes
• Spectra averaging areas increase further out ?
  aggregates more wave information

11
? ? Data Collection

• Data available from Cruise
• Pt Pinos
• Cross Spectra Files Averaged at 10 mins
  intervals
• Wave Files (model/spectra) Averaged at 1 hour
  interval
• Spectral Processing statistics files Averaged at
  1 hour interval
• Used waves files (model/spectra) and statistics
  files
• Wave rider buoy
• Measures wind and swell waves
• Averaged at 30 mins intervals
• Used only on-the-hour data
• Wind Station at Point Pinos (by proximity, though
  winds may be affected by local geography)
• Sampled at 2 secs
• Program to averaged u,v components on the hour,
  with wind data /- 30 mins
• Gives only local winds which drives wind waves
• Is not a good representation of swell conditions
  but a reasonable representation of wind waves

12
? ? ? Data Comparison

• Time Series Comparison

One Glance Appreciation
Model Method
Significant Wave Height
Buoy Wind
Peak Period
Spectra Method
Mean Direction

• Measure of Performance
• RMS errors of HF-derived results from buoy
• No of Hours when results are within tolerance
• lt 10 of buoy Hs
• lt 0.5 sec of buoy Tp
• lt 10 deg of buoy Dp
• Time when all 3 results are within tolerance
  levels

13
? ? ? Data Comparison

• Time Series Comparison

One Glance Appreciation
Model Method
Significant Wave Height
Buoy Wind
Peak Period
Spectra Method
Mean Direction
14
? ? ? Data Comparison

• Time Series

Model Method
Buoy Wind
One Glance Appreciation

• Assuming Buoys wave data is representative of
  ground-truth but
• Short duration peaks 17 sec in earlier phase
• longer duration peaks 18 sec in later periods of
  experiment
• Radar-derived results clustered together
• Radar-derived Hs
• Generally higher than buoys
• Coincides on certain time segments
• Radar-derived Tp
• Hovers around 10 sec
• Coincides on almost half the time
• Radar-derived Dp
• hovers around 270 deg
• Generally negatively biased from buoys except at
  certain segments
• Study periods of coincidence and deviation

15
? ? ? Data Comparison

• Time Series

Model Method
Buoy Wind
One Glance Appreciation

• Wind Speed
• Large variations
• 0.09m/s (_at_ 180100hrs)
• 7.12m/s (_at_ 182200hrs)
• Stronger winds (gt 4m/s) happening at hours of
  darkness
• Wind Direction
• Large variations before 180100hrs
• Predominant SW to NNW after 180100hrs
• SNR of Monopole
• SNR of 3 antenna are closely correlated hence
  choose monopole
• SNR of 3 RCs very close
• Generally gt 30 dB, averaging at 37.5 dB
  throughout (except 1st pt)

16
? ? ? Data Comparison

• Time Series

Model Method
Buoy Wind
Significant Wave Height

• Coincidence of Hs lt 10 of Buoy Hs occurs at
  limited time segments
• wide variation of wind speed
• NW to SW-winds (onshore), except last points
• Higher SNRs (gt30 dB)
• But non-coincidence do occur even if above
  factors are satisfied
• RMS error
• RC1 0.4146 m
• RC2 0.4133 m
• RC3 0.4990 m
• Shoaling effect not manifested in results but RC2
  gives best results

17
? ? ? Data Comparison
Model Method
Buoy Wind
Peak Periods

• Time Series

• Coincidence of Tp lt 0.5 secs of Buoy Tp occurs
  at limited time segments
• wide variation of wind speed
• W-to-NE winds (onshore)
• Higher SNRs (gt30 dB)
• But non-coincidence do occur even if above
  factors are satisfied
• RMS error
• RC1 3.3710 sec
• RC2 3.4632 sec
• RC3 3.4632 sec
• Large RMS errors biased by peaks
• RMS error before 202300h
• RC1 1.6008 sec
• RC2 1.5649 sec
• RC3 1.5888 sec
• Periods relatively unchanged when shoaling

18
? ? ? Data Comparison
Model Method
Buoy Wind
Mean Direction

• Time Series

• Coincidence of Dp lt 10 deg of Buoy Dp negatively
  biased on most time segments
• wide variation of wind speed
• W-to-NW winds (onshore)
• Higher SNRs (gt30 dB)
• But non-coincidence do occur even if above
  factors are satisfied
• RMS error
• RC1 34.9141 deg
• RC2 34.2335 deg
• RC3 39.6934 deg
• Large error probably due to area averaging
• Shoaling/refraction effects somewhat manifested
  in results but RC2 gives best results

19
? ? ? Data Comparison

• Time Series

One Glance Appreciation
Model Method
Significant Wave Height
Buoy Wind
Peak Period
Spectra Method
Mean Direction
20
? ? ? Data Comparison

• Time Series

Spectra Method
Buoy Wind
One Glance Appreciation

• Radar-derived results more random
• 0 values encountered
• Radar-derived Hs
• Generally scattered around buoys
• Segments of Hs 0m
• Radar-derived Tp
• Hovers around 7 to 13 sec
• Segments of Tp 0 sec
• Radar-derived Dp
• hovers either around 270 deg or 080 deg

21
? ? ? Data Comparison

• Waves Spectra

• From 14/07 to 221200hrs
• Peak Energy evident around T 10 sec
• Secondary peak energy at T 11 to 17 sec
• 1.5 order of magnitude energy confined within T
  4 sec
• From 221200hrs to 23/07
• Peak Energy at T 5 sec, decreasing towards T
  6.7 sec
• Secondary peak energy at T 11 sec
• 1.5 order of magnitude energy confined within T
  2.5 sec increasing to T 4 sec

22
? ? ? Data Comparison

• Waves Spectra Comparison

• Energy density spaces increases with range
• Dynamic range of energy density decreases with
  range, i.e. more narrow banded energy available
  for computation
• Spaces corresponded to Hs, Tp and Dp 0 in Time
  Series analysis, i.e. from late 20/7 to early
  23/07
• Weakness of Spectra Method at lower SNR

23
? ? ? Data Comparison

• Time Series

Spectra Method
Buoy Wind
One Glance Appreciation

• Wind Speed
• As previously described
• Wind Direction
• As previously described
• SNR of Monopole
• As previously described

24
? ? ? Data Comparison

• Time Series

Spectra Method
Buoy Wind
Significant Wave Height

• Coincidence of Hs lt 10 of Buoy Hs occurs in
  limited time segments
• wide variation of wind speed
• NW to SW-winds (onshore), except last points
• Higher SNRs (gt30 dB) except at 1st point 0.16
  dB
• But non-coincidence do occur even if above
  factors are satisfied
• RMS error
• RC1 0.4613 m
• RC2 0.6957 m
• RC3 0.7995 m
• RMS error (w/o 0 values)
• RC1 0.4156 m
• RC2 0.5380 m
• RC3 0.5885 m

25
? ? ? Data Comparison
Spectra Method
Buoy Wind
Peak Periods

• Time Series

• Coincidence of Tp lt 0.5 secs of Buoy Tp occurs
  in limited time segments
• wide variation of wind speed
• W-to-N winds (onshore) except last point from S
• Higher SNRs (gt30 dB) except at 1st point
• But non-coincidence do occur even if above
  factors are satisfied
• RMS error
• RC1 5.2765 sec
• RC2 7.9101 sec
• RC3 7.9101 sec
• Biased by 0 values
• RMS (w/o 0 values)
• RC1 3.4945 sec
• RC2 3.0480 sec
• RC3 3.6743 sec
• Periods relatively unchanged when shoaling

26
? ? ? Data Comparison
Spectra Method
Buoy Wind
Mean Direction

• Time Series

• Coincidence of Dp lt 10 deg of Buoy Dp negatively
  biased on most time segments
• wide variation of wind speed
• W-to-NW winds (onshore) except for last point
• Higher SNRs (gt30 dB)
• Bias to near reciprocal bearings
• RMS error
• RC1 366.2977 deg
• RC2 619.1536 deg
• RC3 685.4856 deg
• Large error due to bias
• RMS error (w/o bias)
• RC1 31.0234 deg
• RC2 32.8469 deg
• RC3 41.6106 deg

27
? ? ? ? Discussion

• Measure of Performance Summary
• RMS Error
• No of Hours when results are within given
  tolerance
• Hs lt 10 of buoy
• Tp lt 0.5 secs of buoy
• Dp lt 10 deg of buoy
• Times when all 3 results are within given
  tolerance
• SNR Analysis
• Determine SNR behaviour coinciding with time when
  results coincide

28
? ? ? ? Discussion

• Measure of Performance Summary
• RMS Error

Adjusted to remove 0 values

• Model Extraction technique
• Gives best results with more consistent values
• Not responsive to changes
• Poorest results for Dp as refraction is neglected

• Spectra Extraction technique
• Random values with zeros
• Varies and hence somewhat able to follow ground
  changes

29
? ? ? ? Discussion

• Measure of Performance Summary
• No of Hours when results are within tolerance
  (Compared to total hours 207 hrs)

• Model Extraction technique
• RC2 surprisingly does not give the best results
• Gives better results for both Hs and Tp but
• Poorer Dp results due to consistent
  under-predicting till about 21/7 and
  over-predicting for most part
• Need to conduct SNR analysis

• Spectra Extraction technique
• Unsurprisingly gives worst performance at RC3 due
  to lower SNR
• Surprisingly gives best Dp predictions at RC1
• Need to conduct SNR analysis

30
? ? ? ? Discussion

• SNR Analysis
• Period when results are within given tolerance

(ignore min SNR 0.1614 at 1st point)

• Period when results exceeds tolerance

31
? ? ? ? Discussion

• SNR Analysis
• Period when results are within given tolerance

• Period when results exceeds tolerance

• Hs Model Extraction technique seems to rely on
  SNR

(ignore min SNR 0.1614 at 1st point)
32
? ? ? ? Discussion

• SNR Analysis
• Period when results are within given tolerance

• Period when results exceeds tolerance

• Dp Spectra Extraction technique seems to rely on
  SNR

(ignore min SNR 0.1614 at 1st point)
33
? ? ? ? Discussion

• SNR Analysis
• Period when results are within given tolerance

• Period when results exceeds tolerance

• Tp No co-relation to SNR

(ignore min SNR 0.1614 at 1st point)
34
? ? ? ? Discussion

• Measure of Performance Summary
• Occasions when all three results are within given
  tolerance

• Model Extraction technique
• zero

• Spectra Extraction technique
• 1 incident with below details

15/7 2100 hrs Pt Pinos Wind Speed
Direction 5.92 m/s from 307.9 deg SNR
Antenna 1 34.9 dB 2 36.3 dB 3 39.0
dB Significant Wave Ht (m) HF 1.20 m
Buoy 1.29 m Peak Period (sec)
HF 8.60 sec Buoy 9.10 sec Mean
Direction (deg) HF 298.90 deg Buoy
294.80 deg
35
? ? ? ? Conclusion

• Measure of Performance Summary
• Score Tally

• Model Extraction technique outperforms Spectra
  Extraction technique on most counts except for
  estimating Dp

36
? ? ? ? ? Conclusion Recommendations

• SNR Analysis
• Model Extraction technique
• Relies on SNR for estimating Hs
• Does not seem to rely on SNR for estimating Dp
  Tp
• Spectra Extraction technique
• Relies on SNR for estimating Dp
• Does not seem to rely on SNR for estimating Hs
  Tp
• Relative Strengths of each techniques depends on
  extraction mechanisms
• Model fitting to PM-model
• Spectra model relies on computing Fourier angular
  coefficients from spectra
• Recommendations
• Use Model Extraction technique to compute Hs and
  Tp
• Use Spectra Extraction technique to compute Dp
• Fine-tuning of SeaSonde parameters to process
  waves