1. J. Beaudoin Ocean Mapping Group University of New Brunswick
Real-time Monitoring of Uncertainty due to Refraction in Multibeam Echosounding
J. Beaudoin
Ocean Mapping Group
University of New Brunswick
Oct. 21st 2008
Shallow Survey 2008

2. Introduction CSL Heron CCGS Creed CCGS Matthew CCGS Amundsen 1450 m/s
Oct. 21st 2008
Shallow Survey 2008

3. Assessing Refraction Artifacts in Real-Time
Highly subjective
Requires constant vigil
Can overreact over flat seafloors
Can “underreact” over complicated topography
Impossible for iso-velocity displays (e.g. Reson 81XX display)
Oct. 21st 2008
Shallow Survey 2008

4. Proposed Approach: Raytracing Simulation
Isolates raytracing portion of depth reduction procedure: no sounding data required!
Requires accurate model of raytracing procedure:
Draft
Angular sector
Survey depth
Surface sound speed probe
Sound speed
Depth
TWTT
Investigation depth
Depression angle
TWTT 1 2
Compute common TWTT
Raytrace in test watercolumn
Difference the solutions 3
Depth error
Oct. 21st 2008
Shallow Survey 2008
Horizontal error

5. Application to Real-Time Monitoring
Sound speed
Depth
“Smile” artifacts near surface
Bias cancels out at mid-depth
“Frown” artifacts at depth
It’s important to see the whole picture
Oct. 21st 2008
Shallow Survey 2008

6. Uncertainty Wedge Error (%w.d.) Oct. 21st 2008 Shallow Survey 2008
Sound speed
Depth
Error (%w.d.)
Oct. 21st 2008
Shallow Survey 2008

7. Simplification for Real-Time Decision Making
Sound speed
Depth
% w.d.
% w.d.
% w.d.
% w.d.
> 1.0% w.d.
IHO Order
Allowable depth dependant portion of TVU
Special
0.75% w.d. 1a
1.3% w.d. 1b 2
2.3% w.d.
Oct. 21st 2008
Shallow Survey 2008

8. Snapshots of Refraction Bias Through an Evolving Watercolumn6
5 vs. 6 5
4 vs. 5 1 6
depth
sound speed 2 3 4 5 4
3 vs. 4 3
2 vs. 3 2
1 vs. 2 1 2 3 4 5 6
depth 1
Oct. 21st 2008
Shallow Survey 2008

9. Time Evolution of Bias Between Casts1 2 3 4 5 6
Time Evolution of Bias Between Casts 5- 4 5- 4
depth
sound speed 5 4
HUGE ASSUMPTION: Linear growth of bias with time
Not unreasonable if you’re sampling at a high rate but DEFINITELY not applicable if you’re undersampling
Oct. 21st 2008
Shallow Survey 2008

10. Real-Time Uncertainty Visualization
Look direction
Comparison of cast 1 & 2
Comparison of cast 2 & 3
depth
sound speed 1 2 3
It’s important to be able to visualize the time evolution and history of error
Oct. 21st 2008
Shallow Survey 2008

11. Uncertainty Visualization
Sound Speed Field
Depth
Bottom
Error Analysis
Time
Uncertainty Field
Depth
Oct. 21st 2008
Shallow Survey 2008

12. Real-Time Monitoring
Depth
Time
Oct. 21st 2008
Shallow Survey 2008

13. Other Applications Error Analysis with Raytrace Simulation
Pre-survey Analysis
CSL Heron, Port of Saint John (2008)
Quality Assurance
CCGS Matthew, Advocate Bay (2008)
CCGS Matthew, EM710 acceptance trials (2005)
Oct. 21st 2008
Shallow Survey 2008

14. Example 1: CSL Heron, Port of Saint John
MVP30
Sound speed
Temperature
Salinity
Oct. 21st 2008
Shallow Survey 2008

15. Casts within problem area highlighted
Sound speed field
Problem area
Geographic plot of depth error highlights areas where the watercolumn’s rate of change exceeds our ability to sample it…
HIGH TIDE
Error Field
Casts within problem area highlighted
How does this affect survey planning?
Less of our angular sector is within tolerable uncertainty, so can reduce line spacing in these areas to maintain accuracy
Could reduce vessel speed to increase spatial sampling of the rapidly changing watermass
Could survey at low tide
Error Field
LOW TIDE
Oct. 21st 2008
Shallow Survey 2008

16. Example 2: Post-Survey Quality Assurance
Bay of Fundy
CCGS Matthew
- EM710 (140° sector)
- MVP200
Sound speed
Temperature
Salinity
Oct. 21st 2008
Shallow Survey 2008

17. Post-Survey Quality Assurance
MVP200
Error analysis
- 233 casts over 9.5 hr survey, 2 min. sampling interval
Uncertainty due to refraction maintained within +/- 0.02% w.d. !!
Oct. 21st 2008
Shallow Survey 2008

18. Post-Survey Quality Assurance
Cross sectional view of soundings
Refraction uncertainty is noise in the error budge
Largest source of uncertainty is water level
Soundings shown here are tidally reduced with WebTide (2D barotropic hydrodynamic model)… CHS uses GPS/RTK tide
Oct. 21st 2008
Shallow Survey 2008

19. Example 3: What is oceanographically significant?
2005 CCGS Matthew EM710 Acceptance Trials
MVP200
95 casts collected during transit
Sound speed casts
Only 36 casts required to maintain uncertainty < 0.25% w.d.
Oct. 21st 2008
Shallow Survey 2008

20. How’d you do that?? 95 casts collected 36 casts required
Oct. 21st 2008
Shallow Survey 2008

21. The “Goldilocks” Watermass
95 casts collected
The Oversampled Watermass
36 casts required
The “Goldilocks” Watermass
Oct. 21st 2008
Shallow Survey 2008

22. Conclusion
Ability to monitor watercolumn conditions as a source of error gives unprecedented control over refraction
Surveyors can have confidence in refraction solution in real-time
The ability to “tune” MVP profile sampling rate can minimize wear on equipment while maintaining a desired accuracy: The Goldilocks Watermass
Many other analysis problems are easily solved using the OMG/UNB SVP Toolkit: Pre-analysis, QA
Oct. 21st 2008
Shallow Survey 2008

23. Future Work
Incorporation of UNB uncertainty monitoring in ODIM Brooke Ocean MVP controller
Automated MVP deployment with error monitoring & error prediction
Application of simulator to case of undersampled watercolumn
Oct. 21st 2008
Shallow Survey 2008

24. Acknowledgements NSERC and CFI funding of ArcticNet NCE
Sponsors of the UNB Chair in Ocean Mapping
U.S. Geological Survey
Kongsberg Maritime
Royal (U.K.) Navy
Fugro Pelagos
Route Survey Office of the Canadian Navy
Rijkswaterstaat
Mike Lamplugh & Jon Griffin, CHS Atlantic
ODIM Brooke Ocean
Students of UNB HydroCamp 2008
Oct. 21st 2008
Shallow Survey 2008

25. Oct. 21st 2008
Shallow Survey 2008

26. Extra Slides
Oct. 21st 2008
Shallow Survey 2008

27. Simulation Subtleties
Roll & Pitch
Performance envelope
Along-track slope
Across-track topography
Oct. 21st 2008
Shallow Survey 2008

28. Surface Sound Speed Can mimic use of a surface sound speed probe:
Retrieve sound speed at transducer depth from control cast
Use this to compute ray parameter for raytrace with test cast
Oct. 21st 2008
Shallow Survey 2008

29. Does this actually work?? Refraction Step Artifacts
Step artifact at moment of transition
Along-track view
Across-track view
Depth difference (m)
Across-Track (m)
Difference between swaths before and after transition
Predicted difference from raytrace simulator
Oct. 21st 2008
Shallow Survey 2008

30. What about in between casts?1 2 3 4 5 6
What about in between casts?
Depends if you are sampling the watercolumn at a high rate
Yes?
Are you stuck with real-time reduced soundings (e.g. REA)? ...or…
Can you post-process using “nearest in time”
No?
Hmmmm
Oct. 21st 2008
Shallow Survey 2008

31. depth
sound speed 5 4
Interpolation of Error Between Casts Case 1: Stuck with real-time reduced soundings
Outer beam depth error
0.81% w.d.
0% w.d.
Cast 4
Cast 5
Time
Cast 4 is used up to the moment that cast 5 is acquired
Error is zero at moment just after acquisition of cast 4
Error increases with time (linearly?), reaching a maximum just prior to collection of cast #5
Error returns to zero after acquisition of cast 5, increasing until the next cast
Oct. 21st 2008
Shallow Survey 2008

32. Case 1: Stuck with real-time reduced soundings5- 5 4+
Case 1: Stuck with real-time reduced soundings
depth
sound speed 5 4 4+
Oct. 21st 2008
Shallow Survey 2008

33. sound speed
Interpolation of Error Between Casts Case 2: Will post-process using “nearest in time” 5 4
depth
0% w.d.
0.81% w.d.
Outer beam depth error
0% w.d.
0.81% w.d.
tmidpoint t4 t5
Time
Oct. 21st 2008
Shallow Survey 2008

34. Case 2: Will post-process using “nearest in time”1 2 3 4 5 6
Case 2: Will post-process using “nearest in time” 5- 4+
depth
sound speed 5 4
Oct. 21st 2008
Shallow Survey 2008

35. Case 2: Will post-process using “nearest in time”5 4+
Case 2: Will post-process using “nearest in time”
Error increases with time (linearly?), reaching a maximum at the midpoint between collection of casts 4 & 5; error then decreases with time, reaching a minimum at the moment cast 5 is collected
Oct. 21st 2008
Shallow Survey 2008

36. Case 1 vs. Case 2 Depth Depth Time Last observed in time
Closest in time
Depth
Oct. 21st 2008
Shallow Survey 2008
Time

37. Case 3: Undersampled Watercolumn
Imprudent to interpolate, BUT…
Snapshot of uncertainty is still a useful metric that can be used to compile an “average” uncertainty
ESS: uncertainty of averages
UNB: average of uncertainties
Oct. 21st 2008
Shallow Survey 2008

38. Interpolation of Error Between Casts Case 3: Undersampled Watercolumn
depth
sound speed 16 15
Interpolate error? Bad idea…
Outer beam depth error
Cast 15
Cast 16
Time
Time
Depth
1510 m/s
1450 m/s
Imprudent to interpolate, BUT…
Snapshot of uncertainty is still a useful metric that can be used to compile an “average” uncertainty
Oct. 21st 2008
Shallow Survey 2008