Feature ArticlesUsing GPS-Derived Water Levels For Hydrographic Surveys
By Crescent Moegling
Hydrographic Team Leader
Office of Coast Survey
Pacific Hydrographic Branch
Vice President and Director of
Jason C. Creech
Lead Hydrographer and Project
Marine Services Division
David Evans and Associates Inc.
Director for Marine Products
Richmond Hill, Canada
With a carrier phase-based global positioning system (GPS), hydrographers can now fill in the measurement gap between heave measurements and water level gauge measurements. The effects of squat, settlement and dynamic draft are mitigated.
There are two basic approaches to make use of carrier phase-based GPS: real-time kinematic (RTK) and post-processed kinematic (PPK). These two methods can be further divided into a single-base operation or a virtual base from a network of continuously operating stations. RTK requires the use of a reliable link between the base station and survey vessel.
Post-mission processing removes the effects of gaps in real-time correction reception over radio links or data gaps in general. Although such gaps only account for a small percentage of data acquisition, they must be removed. Post-processing involves the use of base station data recorded on the reference site, which, when combined with raw data recorded on the hydrographic vessel, results in a robust height solution. This allows the hydrographer to replace heave with ellipsoidal height. A smoothed best estimate trajectory (SBET) of position is produced. SBET data provides accurate sub-five-centimeter ellipsoid heights. PPK-based operations have many advantages as well, including eliminating the need for a radio link, the ability to extend the distance from the base to 20 kilometers and improved accuracy, since errors due to GPS outages can be corrected by forward/backward smoothing.
The most recent versions of the Applanix POSPac and other post-processing tools include the post-processed virtual reference station (PPVRS) technique, which makes use of GPS network stations to determine atmospheric biases at rover positions and which tightly integrates GPS with inertial data to provide a continuous, high-precision navigation solution with baselines of as many as 100 kilometers. The expected result is a method for hydrographic survey with the necessary precision for all inland waters and nearshore areas in the continental U.S. Dedicated base stations are no longer required. PPVRS changes the way hydrographic surveys are acquired by reducing logistics and ensuring positioning within five centimeters.
For the survey presented in this article, it was decided to use both RTK and PPK from a single base reference with the base site coordinate in NAD 83 (North American Datum—a geoid model that defines orthometric heights to a reference ellipsoid), although test areas and statistics are provided using the PPVRS option in conjunction with NAD 83 coordinates.
A Case Study: Columbia River
A critical aspect of any hydrographic survey using GPS heights is a separation model from the reference ellipsoid to the chart datum. Detailed models of the relationship between the chart datum and an orthometric height are required to transform ellipsoid heights to water level above chart datum.
The U.S. Army Corps of Engineers, Portland District, provided the profile of Columbia River Datum (CRD) relative to North American Vertical Datum of 1988 (NAVD 88), the vertical control datum established for vertical control surveying in the U.S. The first step in generating the model was to convert the profile down the river into a triangular irregular network (TIN) spatial model of CRD relative to NAVD 88. The river profile was offset at 2,000-foot intervals perpendicular to the profile and modeled down back channels of islands. Data points that define the CRD surface relative to NAVD 88 were converted to NAD 83 ellipsoid heights us-ing the GEOID03 model and converted datum was inserted into the TIN model.
The final step was to convert the TIN model into a high-resolution grid model. The grid model used the same format as geoid grid models generated by the National Geodetic Survey, which can be used in hydrographic software such as HYPACK Inc.'s (Middletown, Connecticut) HYPACK and CARIS' (Fredericton, Canada) HIPS to convert ellipsoid heights directly to a mapping datum. A series of three models were generated at a three-second arc resolution in order to capture the high definition of the TIN model in small channels and at the confluence with the Willamette River.
GPS base stations were established along the project at intervals no greater than 18 kilometers. This allowed for a maximum range from a base station of 10 kilometers. Base stations logged one-second epoch GPS observables and broadcast real-time carrier phase correctors to the two survey vessels every second. A repeater radio was used to relay the signal to reach the 10-kilometer range in some areas. Aboard the survey vessels, the correctors were received and processed by a position and orientation system (POS) for marine vessels with dual-frequency (L1/L2) receivers. As a redundant observation, the correctors were processed with a Trimble (Sunnyvale, California) MS 750 receiver, which provided HYPACK with real-time ellipsoid heights. The CRD separation model was input into HYPACK for real-time water levels.
To verify that the correctors were providing position data within survey specifications, a Trimble DSM 132—receiving differential correctors from the U.S. Coast Guard beacon at Fort Stevens, Oregon—was used for a real-time comparison to RTK position data. Because the survey vessels would switch between base stations, the RTK positions and elevations were logged as the vessel was static, and a comparison was made with the values obtained from each station. Typical comparison values were under two centimeters. A water level-corrected multibeam swath painting in real time, with a data display of real-time water levels and status of RTK and differential positions with real-time coverage display, was color coded for a distinct color change at the two-meter and four-meter depth curves. Survey specifications called for full multibeam coverage to the four-meter curve and 25-meter skunk striping or single-beam coverage inshore of the two-meter curve. Whereas the swath was corrected to chart datum in real time, the colored depth curves could be used during the survey to verify that coverage requirements were exceeded.
Static Vessel Observation. In order to evaluate the performance of GPS water level calculations, a series of one-hour-long survey lines was logged while floating in the immediate vicinity of gauges assigned to the project, including a subordinate gauge set specifically for the project and five Center for Operational Oceanographic Products and Services (CO-OPS) water level gauges. GPS water levels were calculated relative to CRD using the standard project work flow for RTK solutions, single-base inertial-aided kinematic ambiguity resolution (IAKAR) and PPVRS IAKAR.
GPS water levels for each static observation navigation solution were plotted with the water level gauge data and compared. The RTK and the single-base solutions matched well, but there was a notable and unidentified offset in the PPVRS solution of approximately six centimeters. However, though there was a discrepancy in this analysis, prior comparisons between PPVRS solutions and real-time solutions showed agreement tighter than six centimeters.
A statistical analysis of the navigational solutions and gauge output for the static vessel observation shows the GPS standard deviations from gauge values are better than two centimeters for the MS750, POS RTK, POS Single Base and POS PPVRS, although the mean and median are about five centimeters with the application of PPVRS.
GPS Levels vs. Discrete Zoning. In order to evaluate performance of the two methods, a sample data set with depths reduced using GPS water levels was compared to a duplicate data set using discrete zoning with verified CO-OPS water levels relative to CRD.
Comparisons between the two data sets showed differences caused by discrepancies in CRD between the model used to compute GPS water levels and the water level station, as well as differences (errors) caused by the inability of zoning from a stationary gauge to capture water level changes at the vessel. Some water level gauges output water levels using an incorrect CRD-to-NAVD 88 relationship. For areas where significant discrepancies between published CRD (used by the water level gauges) and interim CRD (used in the model) were present, swath to swath agreement between survey lines was evaluated rather than computing a difference surface between the two data sets. Typically, survey line agreement was evaluated between overlap from lines acquired under varying stages of the tide cycle or flow conditions.
Comparisons between adjacent swaths showed differences of more than 40 centimeters with zoning applied and virtually no discernable difference with GPS water levels applied.
GPS-derived water levels provided previously unmeasured water levels at the survey platform and vastly improved vertical accuracy to hydrographic surveys over discrete tidal zoning. Since the phase center of the GPS antenna and the acoustic center of the multibeam sonar are at a fixed distance and move in unison with vessel loading, settlement and squat, vertical uncertainties in static draft and dynamic draft are eliminated.
This article is an abridged form of a paper presented at U.S. Hydro 2009, sponsored by The Hydrographic Society of America. To access the original paper, e-mail firstname.lastname@example.org.
Crescent Moegling is hydrographic team leader for NOAA's Office of Coast Survey, Pacific Hydrographic Branch.
Jon Dasler is the director of marine services and vice president at David Evans and Associates Inc.
Jason C. Creech is lead hydrographer and project manager for David Evans and Associates' Marine Services Division.
Peter Canter is the director for marine products at Applanix Corp.