Optimizing Sound Speed Profiling to Meet TPU Requirements using a CAST Gauge (Computer Aided Sound speed Technology) Derrick R. Peyton, Steven Smyth, Arnold Furlong ODIM Brooke Ocean Jonathan Beaudoin University of New Brunswick Ocean Mapping Group Michael Lamplugh Canadian Hydrographic Service Abstract An important component to determining the Total Propagated Uncertainty (TPU) of a sounding is an understanding of the speed of sound through the water column. If correct modelling of the refraction is not implemented large errors can result in the sounding horizontal position and depth. The introduction of the Moving Vessel Profiler (MVP) has shown that sound speed profiles can be collected at a high spatial and temporal resolution while the survey vessel is under way. However, optimization of the entire multibeam data collection system is required in order to minimize maintenance costs as well as to apply effective survey planning and procedures. Therefore, efficient use of the MVP is necessary to ensure the collected multibeam data is adequately corrected for refraction to the point where customer accuracy requirements are achieved. This paper will describe the implementation of an MVP real-time CAST Gauge (Computer Aided Sound speed Technology) that integrates the science behind the uncertainty wedge and that of the Total Propagated Uncertainty (TPU) with respect to refraction. The CAST Gauge is a visual and numerical tool that will assist the hydrographer in deciding when he/she should make a sound speed cast, how deep the cast should be, and at what times the casts should be executed. All based on the survey requirements. US Hydrographic Conference, Norfolk, Virginia, 2009 1
Introduction All hydrographic surveys need to address accurate measurement of sound speed in the water column. This is particularly true with respect to multibeam surveys where the outer beams are subject to potentially large propagation errors resulting from an incorrect refraction correction due to poor knowledge of the depth, spatial, and temporal variations in the speed of sound. It thus becomes essential that the hydrographer has a good understanding of the oceanographic characteristics of the survey area in order to monitor relevant changes in water column and make the necessary observations in order to minimize errors associated with sound speed variations (refraction). Numerous papers show variations in sound speed are significant and warrant repeated and close observation in order to properly acquire and process multibeam data (Hughes Clarke et al, 2000; Cartwright 2003). Studies have shown that variations in sound speed affect horizontal and vertical position of a sounding by as much as several meters (DaSilva, 2001; Batton, 2004). To statistically address and present hydrographic depth information the IHO Standards for Hydrographic Surveys (Special Publication S-44) requires that hydrographic surveys account for sound speed errors in order to determine the TPU (Total Propagated Uncertainty) as well as THU (Total Horizontal Uncertainty) and TVU (Total Vertical Uncertainty). The CHS and NOAA both stipulate in their field procedures and deliverables that the TPU of a sounding be provided as a deliverable of hydrographic surveys. (NOAA s 2008 Field Procedures Manual; CHS Standards for Hydrographic Surveys). Total Propagated Uncertainty (TPU) it is an estimate of the uncertainty of any individual sounding, taking into account the uncertainty estimates of the component measurements (tide, sound speed, draft, range measurement, angle measurement, attitude, offsets, etc), expressed as a separate value in horizontal and vertical planes. Uncertainty in the measurement of sound through the water column is one of the largest sources of error in modern surveying (Finlayson, 2008). Some work has been accomplished in estimating the sound speed component of TPU (Imahori and Hiebert, 2008). For the most part, charting agency survey requirements place the onus on the contractor (or the hydrographer) to provide all the necessary methods and processes such that the deliverable hydrographic data meets the desired n th Order Specifications. However, it can be argued that in regions of significant bathymetric relief, it is quite difficult to determine from processed data sets if sound speed uncertainties are properly modelled Figure 1 shows the familiar effect of unmodelled sound speed on multibeam data. In this particular instance of 90m water depths, the vertical spread between the outer beam sounding and the nadir beam is more then 3m 1. Observation and analysis of this type of effect is relatively easy when the seafloor is flat. Therefore, the decision to conduct sound speed casts more often is well supported. 1 Data provided by NOAA s Office of Coastal Surveys. US Hydrographic Conference, Norfolk, Virginia, 2009 2
Figure 1: Bathymetric effect of unmodelled sound speed in flat seafloor A complex seafloor is often a more realistic representation of the bathymetric characteristics. In this situation it would be more difficult to determine errors in bathymetry attributable to poorly modelled sound speed information as these could be masked by significant topographic variations. The NOAA 2008 Field Procedure s Manual states; The underlying principle is that the measured sound speed uncertainty depends on the spatial and temporal environmental variability and the frequency at which sound speed casts are taken. Typically, the amount of time required to obtain additional sound speed profiles is far less than that required to edit, or otherwise fix, data afflicted with acoustic refraction artifacts. Without a very good understanding of the oceanographic characteristics of the survey area it is impractical in many cases to assume a limited amount of sound speed profiles will be sufficient to meet hydrographic survey requirements. The hydrographer should not simply plan to fix or flatten the observed bottom. To this end, the ODIM MVP (Moving Vessel Profiler) has been introduced as a means of increasing spatial and temporal sound speed profiles though an integrated and automated vessel underway winching system. To date, nearly 100 of these systems are in use worldwide. However, with any winching system, the MVP is composed of moving parts that, over time, will require maintenance and service. ODIM has observed many users who are either not acquiring enough sound speed profiles, or they are collecting too many US Hydrographic Conference, Norfolk, Virginia, 2009 3
casts, causing unnecessary financial burden on survey operations due to increased maintenance costs associated with heavy usage of the MVP. Real-time Water Column Profiling The ODIM Moving Vessel Profiler (ODIM MVP) greatly enhances the productivity of CTD, Sound Speed and other specialized profiling by allowing water column casts to be conducted from an underway vessel. The ODIM MVP consists of sensors housed in a small, streamlined free fall fish, a conductor cable with strength member, a computercontrolled high speed hydraulic winch and a complete cable metering, overboarding and docking system. The sensor information is transmitted in real-time through the conductor cable to the survey vessel for immediate input into the multibeam collection system (and of course, for data processing). The ODIM MVP allows the free fall fish to fall near-vertically. Deployment is executed under computer control and can be restrained by three parameters: 1) desired depth of cast, 2) preset height above the bottom, or 3) maximum cable out. Using this concept, the system can achieve a much deeper depth for a given vessel speed than a comparable towed system. Once the programmed downcast depth has been reached, the fish is towed near the surface where it can be recovered or redeployed. Figure 2a shows the MVP200 presently installed on board the NOAA Ship Fairweather. Figure 2b provides the Graphic User Interface of the MVP system labelling the relevant integrated system components. Figure 2a: ODIM MVP200 on NOAA Ship Fairweather US Hydrographic Conference, Norfolk, Virginia, 2009 4
Figure 2b: ODIM MVP200 Graphic User Interface The NOAA Field Procedure manual recognizes the advantage gained by sampling the water column at high rates and allows for a fourfold reduction of estimated TPU due to sound speed uncertainty in the case where the water column is sampled at intervals less then 15 minutes from every 4 hours. We note that this approach recognizes that the water column variability is narrowed down to a single uncertainty value. The benefits of increased sound speed profiles are shown by Dr. Jody Klymak, University of Victoria, and illustrated in Figure 3. Dr. Klymak makes an MVP200 CTD cast every 2 nm at 10 knots (i.e. cycle time of <10 minutes) with a payload that includes an SBE49 CTD, an AML O2, and a Wetlabs Fluorometer. The figure shows comparative temperature profiles contrasting the resolutions achievable with an MVP (bottom) versus the traditional CTD (top) profiling approach (Klymak, 2009). Figure 3: MVP200 Casts vs Traditional CTD Casts (Klymak, 2009) US Hydrographic Conference, Norfolk, Virginia, 2009 5
The challenge thus becomes; How can we minimize ODIM MVP operational costs with respect to maintenance and service while at the same time use the merits of the MVP to reduce the sound speed component of the Total Propagated Uncertainty? Real-time Monitoring and Assessment of Sound Speed Refraction artifacts are typically death with in post-processing which is both time consuming and requires significant processing expertise. In addition, under-sampled sound speed profiles can result in non-recoverable errors. One approach to monitoring and assessing refraction artifacts in real-time is to monitor the sound speed variability that is the cause of the artifact. This approach isolates the ray tracing portion of depth reduction procedure and computes the bias in sounding depth and horizontal position that would be incurred had the most recent profile NOT been collected, i.e. the previously collected cast had been used instead (Beaudoin, 2008). As the potential bias can vary dramatically with depth and incidence angle, it is computed over the entire sounding space, from sounder to seafloor and across the entire angular sector. This allows for the creation of a wedge-shaped bias lookup table ( uncertainty wedge ) which can be used to determine which, if any, portions of the mapped sector are suffering from unacceptable biases. By comparing sequential pairs of sound speed casts, the hydrographer can ascertain if the current profiling rate is sufficient for maintaining accuracy. If it is, the operator can lower the rate in order to reduce mechanical wear on the MVP. On the other hand, if the profiling rate is insufficient for maintaining the desired accuracy, the surveyor can react by either increasing the profiling rate or decreasing survey line spacing, or a combination of both. With the uncertainty wedge the hydrographer is acutely aware of the impact sound speed accuracy on survey results. Using this tool to monitor sound speed variability should greatly reduce refraction artifacts in the raw data thus minimizing postprocessing efforts. A visualization of the interrelationship of the uncertainty wedge with survey specifications (e.g. IHO S-44 Survey Specifications) and a multibeam swath is illustrated in Figure 4 (Beaudoin, 2008). US Hydrographic Conference, Norfolk, Virginia, 2009 6
Figure 4: Sound Speed Uncertainty Wedge (Beaudoin, 2008) The MVP CAST Gauge Advantage The benefits of this approach are realized when we integrate the concept to the ODIM MVP technology. The goal is to provide a mechanism that will enable the hydrographer to predict when to make a sound speed cast based on quantifiable information such as the sound speed uncertainty in the water column as observed along a sequence of casts performed by the MVP in real-time. Again, it is clarified here that the uncertainly value in this context is the difference between the truth or epoch 2 sound speed profile, to the sound speed profile of epoch 1. No consideration at this point is given to measurement uncertainty of the sensor itself. Figure 5 shows a sequential set of sound speed casts collected using an ODIM MVP200. Figure 5: ODIM MVP along track Sound Speed US Hydrographic Conference, Norfolk, Virginia, 2009 7
Figure 5 presents the along track sound speed profiles as observed by the MVP. The color intensity plot is an intuitive graphical presentation of sound speed variability over a survey line. The observed sound speed color legend is shown on the right. When set to automatically update, the picture of the developing variability is updated in real-time. Each file represents a cast and each display will represent the accumulated casts taken over a line. Figure 6: ODIM MVP CAST Gauge Figure 6 shows the Uncertainty Wedge integrated with the MVP Controller software to form a CAST (Computer Assisted Sound speed Technology) Gauge for sound speed in the water column. For Figure 6, if profile # 11 is the most recent profile with # 10 being the previous profile, comparison of the two profiles using the Uncertainty Wedge shows that the difference is < 0.75% W.D. for all angular sectors. If a maximum of 1.00% W.D. uncertainty is assumed to be the cut-off, this means that the water column was adequately sampled between epochs #10 and #11. It is hoped that the ODIM MVP CAST Gauge will greatly enhance the hydrographer s knowledge of the oceanographic environment in the survey area. Ultimately this will reduce the overall cost of the survey by reducing the amount of time spent cleansing sound speed artifacts in the multibeam data. Conclusions and Future Work Due to potentially high spatial and temporal variability, the sound speed component of TPU is one of the most difficult parameters to monitor. To reduce this uncertainty it is recommended to increase sound speed profile acquisition rates. This approach is quite costly if it involves stopping a survey vessel. The ODIM MVP provides the advantage of acquiring sound speed profiles while the survey vessel is underway. However, increased use of the MVP means potentially increased costs in maintenance and service. US Hydrographic Conference, Norfolk, Virginia, 2009 8
The software behind the MVP CAST Gauge (Computer Assisted Sound speed Technology) has been developed to optimize maintenance cost and at the same time monitor sound speed uncertainty. The CAST Gauge integrates the Uncertainty Wedge concept of computing and visualizing sound speed uncertainty by comparing ray path analysis from one epoch to the next and displaying this result in real-time. In this way the hydrographer can execute sound speed casts based on a near real-time quantitative analysis of sound speed uncertainty and thus optimize the cost of maintenance and service on the MVP winch; as well the prime outcome which is a hydrographic survey that meets desired accuracy standards. Future work in this area centres on the integration of historical sound speed information with observed MVP casts. The historical data would involve using such databases such as the World Oceanographic Database (WOD), the World Ocean Atlas (WOA), and the Generalized Digital Environmental Model (GDEM). There are additional plans to develop an automated sampling algorithm which adapts sound speed environment in the survey area. Based on historical and predicted sound speed variability, the MVP control system will not only determine how often to sample but also how deep to sample. This satisfies the goal of minimizing operations of the mechanical winching system. In addition, it would be desirable to investigate the 3-dimensional aspect of the monitoring sound speed uncertainty. This work would involve integrating along track and adjoining track sound speed profiles into a 3-dimensional environment. Modelling the 3D environment would lead towards more accurate ray tracing results and thus a more accurate uncertainty wedge. The benefits are not only foreseen in shallow water surveys, where sound speed uncertainty is critical in the outer beams, but also in deep water surveys where, in many cases, a high geospatial resolution of sound speed profiling is not believed necessary. A recommended direction for further study would be to include tidal and current influences and thus working towards a 4D modelling environment. References Batton, D., The Effect of Refraction on Oblique Angles of Multibeam Echo-Sounders due to Sound Speed changes through the Water Column, The Hydrographic Journal, No. 113, July, 2004, page 15-20. Beaudoin, Jonathan, Real-time Monitoring of Uncertainty due to Refraction in Multibeam Echosounding, Paper 10, Shallow Survey 2008, Portsmouth, New Hampshire, USA. Cartwright, D. and John Hughes Clarke, Multibeam surveys of the Fraser River Delta, Coping with an Extreme Refraction Environment, Proceedings of the US Hydrographic Conference, San Diego, USA, 2005. Canadian Hydrographic Service, Standards for Hydrographic Surveys, Department of Fisheries and Oceans, www.charts.gc.ca/pub/en/survey/intro.asp. US Hydrographic Conference, Norfolk, Virginia, 2009 9
DaSilva, Jana L., Contract Hydrography: An Opportunity for Innovation in Hydrographic Surveying, U.S. Hydrographic Conference 2001, May 22-24 Finlayson, David P., Modern Sonar, Washington State Mapping Workshop, 2008. Imahori, Gretchen, and James Hiebert, An Algorithm for Estimating the Sound Speed Component of Total Depth Uncertainty, Proceedings of the Canadian Hydrographic Conference and National Surveyors Conference, 2008. International Hydrographic Organization, IHO Standards for Hydrographic Surveys; Special Publication S-44, Monaco, www.iho-ohi.net/english/standardspublications/catalogue.html#s44. Klymak, Jody, Hi Res Sampling on Line P, University of Victoria, Presentation at the Institute of Ocean Sciences, Sydney, British Columbia, Canada, 2009. Office of Coast Surveys, Field Procedures Manual, NOAA, www.nauticalcharts.noaa.gov/hsd/fpm/fpm.htm US Hydrographic Conference, Norfolk, Virginia, 2009 10