Lawrence H. Haselmaier, Jr. Naval Oceanographic Office U.S. Hydrographic Conference February 2015

Transcription

1 Lawrence H. Haselmaier, Jr. Naval Oceanographic Office U.S. Hydrographic Conference February 2015 Adding an Ellipsoid Reference to AUV Multibeam Data to Meet Hydrographic Standards Abstract: One challenge for Autonomous Underwater Vehicle (AUV) multibeam surveying is the limited ability to assess internal vertical agreement rapidly and reliably. Applying an ellipsoid reference to AUV multibeam data would allow for field comparisons; however, traditional methods are precluded due to the absence of a continuous source of ellipsoid height (EH) data from the AUV. Data collected by a surface vessel in close proximity to the AUV could provide the necessary EH source. During three AUV multibeam missions, a boat stayed near the AUV and collected EH data. Virtual tide corrector values for the AUV multibeam data were derived using EH data and a measured ellipsoid to vertical datum separation value. Those virtual tide corrector values were compared to measurements by a National Oceanic and Atmospheric Administration tide gauge installed nearby and found to agree within 0.12 meters at a 2-sigma confidence interval. Further testing is planned to identify and understand sources of error and to minimize overall uncertainty in the process. Lawrence H. Haselmaier, Jr. Oceanographer Hydrographic Department Naval Oceanographic Office 1002 Balch Blvd. Stennis Space Center, MS (228) lawrence.haselmaier@navy.mil Mr. Haselmaier is an Oceanographer for the Naval Oceanographic Office s Hydrographic Department. He has received his Bachelor of Science in Physics from the University of Alabama. He is currently studying Applied Physics with an emphasis on signal processing at the University of New Orleans. The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of the Department of the Navy, Department of Defense, nor the U.S. Government. Distribution Statement A: Approved for public release; distribution is unlimited. 1

2 Adding an Ellipsoid Reference to AUV Multibeam Data to Meet Hydrographic Standards 1. Introduction Autonomous Underwater Vehicles (AUVs) are an increasingly important tool for the collection of hydrographic data (Hiller, Steingrimsson, and Melvin 1). While the benefits of using AUVs for data collection are numerous, they are of little importance if the data collected have too large a positional uncertainty to meet the requirements of the hydrographic products they will be supporting. Tide offsets are one of the largest sources of uncertainty in hydrographic data, and although traditional observation of tidal fluctuation near the survey area can account for changes and reduce this uncertainty, operational constraints often prevent continuous and local tidal observations (Brennan et al. 1). Applying an ellipsoid reference to AUV multibeam data would achieve the same purpose without the need to collect observed tides, but a significant challenge exists in that an underwater vehicle cannot collect ellipsoid height (EH) data directly. These data could be collected from a nearby surface vessel and then be used to derive virtual water level values, thereby bridging the gap between requirement and achievability. This experiment examines the viability of a method for generating virtual water levels using EH collected from a nearby surface vessel. 2. Proposed Solution Applying a water level corrector derived from EH measurements to AUV multibeam data could result in significantly lower overall vertical uncertainty compared to applying water level correctors from predicted tides. It would also reduce the required data collection effort compared with applying corrections via observed tides. In particular, shore-based gauge installation would not be necessary during data collection activities except as required to validate modeled values 2

3 for the ellipsoid to chart datum separation (SEP) (Dodd et al. 12). Although determining the SEP requires a combination of measurement and modeling, such determination can be made once and used again throughout subsequent surveys in the same area. Depending on the requirements driving the multibeam data collection, it is possible that only relative vertical alignment of multiple survey lines is essential. In that case, any SEP model of reasonable quality would provide the input needed to obtain relative vertical alignment. Once the SEP is known, the chief requirement of deriving virtual tide corrector values is a continuous source of EH data. For surface vessels equipped to collect Global Navigation Satellite System (GNSS) data, this source is readily available. Because it is a submerged vehicle, using on-board GNSS collection capabilities is not an option. However, if a surface vessel were to remain sufficiently close to the horizontal position of the AUV such that the SEP value for both locations was approximately the same, then that surface vessel could supply the continuous EH source for this technique. The surface vessel could be a ship, a small boat, a GPS buoy, or an autonomous surface vehicle (ASV). For the purpose of this study, Meriel B., a 50-foot work boat owned by Hydroid and built by Millennium Marine, was chosen as the surface vessel. Figure 1 depicts the derivation and makeup of a conventional tide corrector for AUV multibeam data. 3

4 Figure 1 Diagram of conventional AUV water level correction with vertical locations of Tide Corrector, Depth Corrector, Raw Sounding, and Corrected Sounding indicated. Figure 2 depicts the derivation and makeup of a virtual tide corrector for AUV multibeam data. Note that the terms contributing to the corrected sounding (highlighted in purple) are equivalent to the virtual tide corrector which contributes to the corrected sounding in the same way as the conventional tide corrector depicted in Figure 1. 4

5 Figure 2 Diagram of virtual AUV water level correction with ellipsoid in green and vertical locations of Tide Corrector, Depth Corrector, Raw Sounding, Corrected Sounding, Ellipsoid Height, Antenna Height, and Ellipsoid to Chart Datum Separation indicated. To determine the fitness of the virtual method for correcting AUV multibeam data, the virtual data should be compared with observed, conventional tidal elevations collected in the same area. If the virtual method can provide data with similar reliability to directly observed water levels with less constraining collection requirements, it will prove useful in helping AUV data meet hydrographic standards. If the data are determined to be reliable, it will be fairly straightforward to incorporate that data into any standard workflow for vertical correction. 5

6 3. Method 3.1. Data Collection A C-Nav3050 GNSS receiver was installed in the cabin of Meriel B. The receiver was connected to a cabin-top-mounted NavCom Technology, Inc. ANT-3001R rover antenna measured at 3.78 meters above the waterline, 0.60 meters port of amidships, and 1.45 meters forward of the rear edge of the cabin top. These measurements were obtained using a tape measure and are expected to be accurate within 3 centimeters. Antenna placement was chosen to maximize satellite visibility without the need for special equipment or undue human risk during installation. Mounting the antenna in a temporary fashion allowed for efficient installation and breakdown while limiting movement of the antenna to less than 1 centimeter in any direction. The reductions in antenna height above the waterline caused by roll, pitch, and settlement were neglected for this experiment and are accounted for in the error analysis portion of this paper. The horizontal displacements arising from roll, pitch, and yaw effects are sufficiently small in comparison with the horizontal distance between Meriel B. and the AUV such that they were disregarded for this experiment. A Remote Environmental Monitoring Units (REMUS) 600 AUV collected multibeam data near New Castle, NH, while Meriel B. remained within approximately 1 kilometer of the REMUS AUV. Data collection occurred on three consecutive days from August 4 to 6, The REMUS AUV collected Kongsberg EM 3002 multibeam sonar data as well as GNSS, inertial, and physical oceanographic supporting data. Operations during August 4 through 6 consisted mostly of standard bathymetric surveys (i.e., the survey lines were designed to achieve or approach full coverage while minimizing redundancy). Navigation data were collected at 1- second intervals in the C-Nav3050 receiver extending at least 30 minutes before and after 6

7 multibeam data collection. The receiver was turned on and off pierside, so the data collection always included transit legs to and from the survey areas. Navigation and water level data were collected for 6, 8.5, and 7 hours on each of the three survey days, respectively. Weather conditions were largely calm for all data collection periods. A National Oceanic and Atmospheric Administration (NOAA) tidal gauge (Station ID ) collected data throughout the surveys and provided preliminary water elevation data at 6-minute intervals with the exception of two data outage periods. The NOAA gauge stopped recording from 2:30 PM (all times UTC) through the end of the survey day on August 5 and from 1:30 PM to 2:30 PM on August 6. No preliminary observed tides were available for these times; however, verified data from all times eventually became available on NOAA s Tides and Currents website. Observed, verified tides for all times were downloaded on February 9, During the first two days, the survey areas were 11 to 13 kilometers east of the tide gauge. During the third day, the survey area was 1 to 2 kilometers east of the tide gauge Data Processing Raw GNSS data collected by the C-Nav3050 receiver on Meriel B. were downloaded daily. The GNSS data were combined with clocks and rapid ephemerides data from the University of Bern s Center for Orbit Determination in Europe (CODE) and were processed using Waypoint GraphNav to provide a Precise Point Positioning (PPP) solution of antenna height with respect to the ellipsoid and antenna geographical location. For this experiment the geographical portion of the PPP solution was neglected. To account for the measured height of the antenna above the waterline, the height data were further reduced by 3.78 meters. 7

8 A program utilizing a discrete Fourier transform was developed in order to account for the rapidly changing motion of the vessel. The frequency domain data were multiplied by the following function to low-pass filter the data: H(s) = N sin(s π L N ), where H(s) is the filter function in the frequency domain, N is number of EH samples, and L is filter length in number of samples. The value for L was chosen to be 60 samples (or 1 minute for this 1-Hz data) to both preserve effects on EH arising from long-term tidal changes and remove rapid changes in movement of the vessel. An inverse discrete Fourier transform was then applied to the data to return it to the time domain. Filtration in the frequency domain was chosen based on convenience and straightforwardness vice time domain filters (Press et al. 558). The virtual tide corrector was calculated for each waterline-corrected EH by taking the difference of the height and the ellipsoid to the Mean Lower Low Water SEP in the area. Byrne, et al. used NOAA s Horizontal Time-Dependent Positioning tool (HTDP) to compute the SEP to be meters at the NOAA tide gauge during work completed during the Shallow Survey 2008 conference (9). That benchmark information was processed using Leidos Survey Analysis and Area-Based Editor (SABER) and combined with the National Geospatial-Intelligence Agency s Earth Gravitational Model 2008 (EGM2008) to generate a calibrated SEP grid for the area. Table 1 describes the positions and SEPs for the NOAA gauge and for the areas surveyed during this experiment. Each SEP was used for virtual tide corrector computation for data collected in the corresponding area. L s π 8

9 Table 1 Locations, positions, and SEPs of NOAA tide gauge and daily survey areas Location Latitude Longitude SEP (m) NOAA Fort Point Tide Station ' N ' W August 4 Survey Area ' N ' W August 5 Survey Area ' N ' W August 6 Survey Area ' N ' W For each 6-minute interval where data existed for both the NOAA tide gauge and the C-Nav3050 receiver, the reduced and filtered EH measurements were averaged over that 6- minute period (e.g., all EH measurements from 12:33 to 12:39 PM were averaged to obtain the virtual tide corrector for 12:36 PM). 4. Results For each day of data collection, the difference between each direct water level measurement from the NOAA gauge and the virtual tide corrector for the same time period was obtained. Both data series and their differences are depicted in Figures 3, 4, and 5 for each day of data collection. 9

10 Figure 3 August 4, 2014 comparison of conventional and virtually determined water levels. Mean difference = 4.7 cm, median difference = 4.9 cm, and standard deviation of differences = 4.2 cm. 10

11 Figure 4 August 5, 2014 comparison of conventional and virtually determined water levels. Mean difference = 8.9 cm, median difference = 8.2 cm, and standard deviation of differences = 7.6 cm. Arrow denotes time after which the NOAA gauge showed no preliminary data collection (2:30 PM). 11

12 Figure 5 August 6, 2014 comparison of conventional and virtually determined water levels. Mean difference = 6.9 cm, median difference = 5.3 cm, and standard deviation of differences = 7.4 cm. Arrows denote start and end of period during which the NOAA gauge showed no preliminary data collection (1:30 PM and 2:30 PM, respectively). Overall, 226 conventional water level data points were compared with corresponding virtually derived measurements. The virtually derived correctors were larger (which would result in shoaler-corrected multibeam data) by a mean of 7.1 centimeters with a median difference of 6.6 centimeters and a standard deviation of 7.0 centimeters. If the original periods of NOAA gauge inoperability are removed, 175 data points remain with a mean difference of 5.4 centimeters, a median difference of 5.0 centimeters, and a standard deviation of 5.9 centimeters. 12

13 August 4 saw virtual tide correctors that fluctuated from very close agreement with NOAA gauge data to up to values larger by up to 13 centimeters. The periods of discrepancy for this day appear to coincide with periods where Meriel B. was operated at the highest speeds. The effects of speed on dynamic draft were neglected for this experiment but could explain erroneously high virtual tide correctors for those periods. Another period of high discrepancy in the results are from August 5 from 4:30 PM through the conclusion of that day s data collection at approximately 8:30 PM, at which time the virtually derived water levels averaged 16 centimeters higher than gauge values. Unlike in the August 4 data, these largest deviations do not correspond to periods of fastest vessel speed. For this time period, the gauge data would agree much more consistently with the virtual data if either the virtual data were reduced by 16 centimeters or the gauge data were advanced by 18 minutes. If only the data from August 5 before this discrepant period are considered, the remaining 47 points have a mean difference of 3.0 centimeters, a median difference of 3.2 centimeters, and a standard deviation of 4.7 centimeters. At this time, it is unclear why the discrepancy exists. The data from August 6 were much more consistent and in agreement with measured data. Although there was originally a 1-hour period of gauge inoperability, it was small with respect to the original gap in the August 5 data and did not seem to affect the data quality appreciably. A greater deviation from gauge values was observed for the beginning of the data collection period, and that interval corresponds with somewhat faster speeds for Meriel B. This correspondence suggests that dynamic draft may have played a significant role in this day s deviation, as well. If only the data from August 6 outside of the period of gauge inoperability are 13

14 used, the remaining 65 points have a mean difference of 6.5 centimeters, a median difference of 4.9 centimeters, and a standard deviation of 7.4 centimeters. Overall, the persistently positive mean and median difference values suggest a tendency of one or more measurements to place the vessel s waterline too high, the gauge measurement of the water level too low, or the magnitude of the SEP value too small. Since the gauge measurements and SEP value are outside of the scope of this experiment, future work will attempt to account for the factors that place the vessel s waterline too high. The 2-sigma value for all data (excluding original periods of gauge inoperability) of 12 centimeters suggests that the method of developing virtual tide correctors from nearby EH measurements is a viable one and that the accuracy of the method could significantly benefit from reducing input uncertainties. Figure 6 depicts the statistical information for the difference values from all relevant datasets. 14

15 Figure 6 Statistical summary for difference values in all datasets The largest contributors to a bias are expected to be offset measurement error (both antenna to reference point and reference point to waterline) as well as settlement and squat effects. Increased standard deviation is expected to arise from settlement and squat effects combined with roll, pitch, and heave effects. It is recommended that future testing occurs with a vessel with professionally surveyed offsets, a robust motion sensor, and known settlement and squat values. 15

16 5. Future Work Although this method appears conceptually suitable for providing water level correctors for AUV multibeam data, it would benefit from rigorous quantification as to its accuracy. The limitations of this experiment for the vessel of choice, the proximity of the vessel to the tide gauge, and the time allocated to data collection should be considered when designing future testing to further develop and characterize this method. The most productive change to the experiment involves vessel choice. Ideally, the vessel will have known settlement and squat characteristics (or be stationary), professionally surveyed offsets, and a robust motion sensor. A hydrographic survey launch (HSL), GPS tide buoy, or an ASV would readily meet those characteristics. It would be especially useful if the vessel were already equipped for hydrographic survey (i.e., it had GNSS navigation, multibeam sonar, and motion sensing capability). In that case, corrected multibeam data from the AUV could be compared to data collected by the vessel for vertical alignment. Either an HSL or an ASV would provide the capability to manipulate the horizontal distance between the vessel and the tide gauge to fully examine the distance s effects on the accuracy of the method. A simplified procedure for conducting such a test would be to collect GNSS data at regular intervals of distance away from the tide gauge under otherwise similar environmental conditions and examine the resultant accuracies. Once the bias and standard deviation were minimized and quantified using the aforementioned modifications, the next step in the method s development would be to transform the calculated virtual tide corrections into a tide correction time series and apply it to AUV multibeam data. It is at this stage of development that a thorough examination of input and 16

17 output uncertainties should be undertaken. The data would then undergo comparison with traditionally tide-corrected AUV data to examine the relative amount of vertical offset internally (within the virtually corrected dataset only) and externally (compared with the traditionally corrected dataset). It will be important to begin with a small distance (within or near 1 kilometer) between the AUV and the surface vessel to isolate effects unrelated to that distance. If that testing shows practical reliability of the method, analysis can be expanded to include greater distances to potentially reduce operational constraints for the method. 6. Conclusion This experiment has demonstrated the reliability of the method for deriving virtual water level measurements using EH data collected from surface vessels. If further testing is conducted to minimize and quantify uncertainties, it is feasible that a fully developed method could ultimately lead to the inclusion of AUV multibeam data in traditional hydrographic products without the unwanted constraints of conventional and local tidal observation. More investigation is required to determine the full impact of the method on hydrographic collection capabilities throughout the oceanographic community. 17

18 Works Cited Brennan, Richard, Kurt Hess, Lloyd Huff, and Steve Gill. The Design of an Uncertainty Model for the Tidal Constituent and Residual Interpolation (TCARI) Method for Tidal Correction of Bathymetric Data. Paper accepted for publication in the proceedings of U.S. Hydro 2005, March 29-31, San Diego, California. Web. February 10, Byrne, Shannon, Walter Simmons, Gail Smith, and Bill Mehaffey. A Demonstration of GPS Based Vertical Control, Unaided by a Shore Station. Presentation of Shallow Survey 2008, October 21-24, 2008, Portsmouth, New Hampshire. Web. February 10, Dodd, Dave, Jerry Mills, Dean Battilana, and Michael Gourley. Hydrographic Surveying Using the Ellipsoid as the Vertical Reference Surface. Facing the Challenges Building the Capacity: Proceedings of FIG Congress April 11-16, 2010, Sydney, Austrialia. Web. February 10, Hiller, Thomas, Arnar Steingrimsson, and Robert Melvin. Positioning Small AUVs for Deeper Water Surveys Using Inverted USBL. Taking Care of the Sea: Proceedings of Hydro12. November 12-15, 2012, Rotterdam, the Netherlands. Rotterdam: Hydrographic Society of Benelux, Print. Press, William, Saul Teukolsky, William Vetterling, and Brian P. Flannery. Numerical Recipes in C. Cambridge, United Kingdom: Cambridge University Press, Print. 18