December 1, 2011
This article is a preview of a story appearing in POB's December issue.
As part of its mission to “define, maintain and provide access to the National Spatial Reference System (NSRS),” the National Geodetic Survey (NGS) issued a bold set of milestones for modernizing the NSRS in its 10-year plan, which was laid out in 2008. One of these milestones was the replacement of the North American Vertical Datum of 1988 (NAVD 88) with a new vertical datum in which the zero elevation surface will be defined through a gravimetric geoid model and that will be accessed primarily through GPS and that geoid model. To prepare for this switch, NGS engaged in a new project of gravity data collection and validation, known as Gravity for the Redefinition of the American Vertical Datum, or GRAV-D. Although the 10-year plan targeted 2018 as the completion date of GRAV-D, fiscal and scheduling realities now point toward a 2022 target date.
In order to assure the highest accuracy in the new vertical datum, NGS stated that “the gravimetric geoid used in defining the NSRS should have an absolute accuracy of 1 centimeter at any place and at any time.” Whether this objective is truly achievable remains an open question. However actual answers about the achievable accuracy of a gravimetric geoid model must be provided. For this reason, NGS decided to embark on the first survey (of a likely trio of surveys) specifically intended to evaluate gravimetric geoid model accuracy. This first survey is known as the Geoid Slope Validation Survey of 2011, or GSVS11.
Although the concept of sea level as a zero height surface stretches back into pre-history, development of a formal theory on the computation of the geoid (from gravity measurements) began in earnest in the mid-1800s. In the century and a half since then, gravity has been measured worldwide, theories have improved, and the overall accuracy of geoid models has increased by many orders of magnitude. In fact, in only the last few decades, theoretical geoid accuracy has gone from “less than 1 meter” to approaching 1 centimeter.
However truly achieving any level of accuracy requires independent validation. For this reason, NGS began making plans in 2010 to perform the first geoid accuracy vaildation survey to support the policy of moving to a gravimetric geoid model as the zero elevation surface for the next national vertical datum.
Numerous technical and theoretical issues arose in the initial planning phases. For example, comparison of one gravity-based geoid model to another is not truly an independent validation when they rely on the same gravity data sets. However, the launch of the GRACE (Gravity Recovery and Climate Experiment) and GOCE (Gravity field and steady-state Ocean Circulation Explorer) satellites, provided new, independent, satellite-only models of the geoid, with accuracies of geoid differences at about 1 centimeter over distances greater than about 200 kilometers (GRACE) or 100 kilometers (GOCE). NGS considered that a survey at least 200 kilometers long could be compared to these models as a starting point.
Those familiar with NGS’s work on geoid modeling since GEOID96 might wonder why NGS didn’t immediately default to using the familiar “N=h-H” equation for validating the geoid. Unfortunately, simply removing orthometric heights (H) from ellipsoid heights (h) as has been done in the past fails to address two unresolved issues: 1) that geodetic leveling requires at least one “known” point in order to determine absolute orthometric heights, and 2) the available data for such comparisons in the United States is in NAD 83 (North American Datum of 1983) and NAVD 88, both of which have known nationwide and regional biases at their very definitional level (which is why the NGS 10-year plan calls for replacing both). The second issue can actually be solved by processing GPS data (for ellipsoid heights) and leveling data (for orthometric heights) without resorting to pre-published NAD 83 or NAVD 88 control. The first issue, however, remains an unsolved difficulty when attempting to validate absolute geoid heights from leveling and GPS.
In accordance with the above limitation, the plan for this first survey was to set aside absolute accuracy for the time being and concentrate on differential accuracy. Thus, it was decided to make use of entirely new GPS-derived ellipsoid height differences and leveling-derived orthometric height differences (completely independent of NAD 83 and NAVD 88) in order to compute geoid height differences (“geoid slopes”) to independently compare to those geoid height differences in the gravimetric geoid model.
However, “leveling without gravity measurements…is meaningless from a rigorous point of view.” And while there certainly exists much historic surface gravity in the U.S., much of it is unreliable. Therefore, it was decided early on that the only way to properly use new leveling to get these orthometric height differences as accurate as possible was to acquire new terrestrial (absolute and relative) gravity measurements at the same time.
Because accuracy of geoid slopes was chosen as the goal of this proposed survey, one additional method of determing accurate geoid slopes was proposed–a survey of astro-geodetic deflections of the vertical (DoV). This is because, at the surface of the geoid, deflections of the vertical are identical to geoid slopes. At the surface of the Earth (where DoV measurements are actually taken), one must make certain calculations to account for the curvature of the plumb line between geoid and Earth’s surface in order to compare to actual geoid slopes. Unfortunately, NGS had last performed regular field surveys of this quantity back in the late 1970s as part of the development for NAD 83 (Dracup, 1994). Thankfully geodetic work on both the theory and instrumentation of determining DoVs was still being performed elsewhere, particularly in Europe. Of particular interest was the work in Switzerland using newly designed camera/GPS systems, such as the DIADEM camera. Because DoVs provide a third independent look at geoid slopes (different from satellite-only models or GPS/leveling derived models), it was decided that NGS should attempt to include them if at all possible.
Other datasets were considered for collection, ranging in categories of importance from “very important” to “interesting, but unnecessary.” Among these were a new digital elevation model (DEM)(very important for processing leveling, DoVs and improving gravimetric geoid models), borehole samples of rock density between the Earth and the geoid (useful, but expensive), borehole gravity (useful, but expensive) and new imagery (useful, but not required for any processing). In the end, a new LiDAR-based DEM was put into the plan, but no other additional datasets were used.
When NGS got serious about these plans in mid-2010, it became necessary to choose exactly where such a survey would take place, how big it would be and when it would be performed. In order to make this choice, some disparate variables had to be considered. It was agreed that the survey should:
1. Fall completely under already-flown (or very soon to be flown) GRAV-D flights to enable NGS to check the accuracy of gravimetric geoid models with and without the new airborne gravity.
2. Be at least 200 km long in order to compare against GRACE and GOCE
3. Be low in absolute elevation, minimizing the assumptions about density of rock and curvature of the plumb line between the earth’s surface and the geoid
4. Not have any significant water crossing (for leveling)
5. Have relatively open sky views (for GPS)
6. Be during a time with minimum cloud cover (for DoVs)
7. Have some existing geodetic marks
8. Run along a roadway
9. Be no sooner than the middle of 2011 (due to availability of the DIADEM camera)
10. Be in a location that is not expensive to travel to
11. Have all the various observation types occur as near in time as possible
At the time of planning, the only significant areas of GRAV-D coverage (requirement No. 1) were Alaska, Puerto Rico and the Gulf Coast of CONUS. The first two areas were considered too expensive (requirement No. 10), and so the Gulf Coast was chosen. A number of line configurations were looked at, though none was ideal (requirements 1, 4, 5, 7 and 8 being the most frequent difficulties). The closest line to ideal was identified as a 330 kilometer (205 mile) line running from the Rockport Tide gage (near Corpus Christi) to the Stephen F. Austin building in Austin, Texas. This line was selected, although parts of it did have some moderately high elevations (maximum of about 250 meters). Both to ensure DIADEM camera availability as well as minimum cloud cover, it was decided to perform the survey in the late summer into the early fall of 2011. Unfortunately, locking in that timeframe so far in advance did not allow for any schedule flexibility when a massive heat wave struck Texas in July and August of 2011 andwildfires plagued the region in September, making working conditions more difficult and dangerous than usual.
Once the line was chosen, reconnaissance to find existing geodetic marks began. The recon crew arrived in Texas in February 2011. The initial results were disheartening. A 25-kilometer stretch just south of Austin had never been surveyed by NGS and therefore had no marks to be found. However, that stretch of road was under construction, and the survey control marks for that construction project turned out to be perfect for the NGS team’s use. The rest of the line contained about 300 NAVD 88 bench marks, but this wasn’t good news since NGS’s standard operating procedure for installing such marks had been to place them either on large concrete structures (the edges of bridges, culverts, etc.) or right next to poles (telephone, fence, etc.) to maximize their stability and/or minimize their disturbability. Nearly every mark found was compleltely unsuitable for a GPS tripod, a gravimeter or the DIADEM camera or, most often, all three.
Along the entire 330 kilometers, only 23 existing marks were deemed suitable for leveling and GPS, and still these marks required the installation of a concrete “collar” to provide a platform for the A-10 gravimeter. Unfortunately, these 23 marks were believed to be unsuitable for the camera (whose standard survey configuration is to be towed on a roadway trailer over geodetic marks). However, an alternate procedure for measuring DoVs with the DIADEM camera had already been discussed and agreed to in order to accommodate marks that couldn’t be driven over.
Mark setting began in earnest in May 2011, with the intent of installing new geodetic marks that were accessible to level rods, GPS fixed height pole tripods, the A-10 absolute gravimeter, a relative gravimeter and the DIADEM camera. Most new marks were concrete pillars with a survey disk, but about 5 percent were a sleeved rod driven to refusal, installed so that repeat surveys on this line would be able to distinguish between near-surface and deep subsidence.
At the end of recon and mark setting, a set of 218 “official” marks were chosen along the route, spaced an average of 1.50 kilometers (0.93 miles) from one another. Of these, 182 were newly installed NGS marks, 23 were existing NGS marks and 13 were marks recently installed by Central Texas Highway Constructors LLC as road construction survey control between Austin and Lockhart Texas and deemed suitable for use in this project.
NGS maintains field crews for both operational work on airport surveys as well as experimental surveys. However, a survey of this scope (spatially, temporally and instrumentally) had not been attempted in some time. Thankfully, NGS did have institutional knowledge and a cadre of employees interested in temporarily getting some field experience. Between full-time field crew members and the training of volunteers, NGS was able to schedule field work for all aspects of the entire survey. Two leveling crews, a GPS crew, a gravity crew,a camera crew and a LiDAR/imagery crew were all identified, with NGS employees rotating in over the course of a planned 11 weeks.
Initially, two leveling crews (eventually increased to four crews) of four people each performed geodetic leveling over all 218 official marks, as well as making regular ties to existing NAVD 88 bench marks along the line. The crews followed procedures for first order Class II leveling and used three different instruments (a Leica DNA03 and a Trimble DINI 11, which was later replaced with a DINI 12). Leveling began on July 18 in Austin, and crews immediately were contending with regular daily high temperatures over 100 degrees Fahrenheit.
During the initial two weeks, some experimental field procedures were investigated. Rather than performing a simple backsight/foresight (BF) at each instrument setup, a modification, with the intent to catch blunders and small systematic errors, was tried. The procedure, called a BFFB, involved taking a backsight/foresight on odd setups. The setup was then off-leveled and re-leveled, and the crew took a new set of foresight/backsight. On even setups, the order was FB-offlevel-relevel-BF.
Unfortunately, the length of time necessary to collect this data conflicted with the schedule for the survey. The experiment was abandoned with the intent that it would be revisited later in 2012 during milder temperatures and under more controlled conditions.
GPS played a role in a variety of components of GSVS11. The most prominent was that a static GPS survey was performed on each of the 218 official marks. This was done with 20 receivers at a time, running approximately 20 hours simultaneously. Each day, the rear 10 receivers were leapfrogged over the front 10 so that another 20-hour dataset could be collected. This process was repeated until the entire set of points had at least two 20-hour GPS data sets.
Because the goal of this component of the survey was accurate differential ellipsoid heights, a number of non-standard procedures were followed. First, in order to maintain constant satellite constellation geometry, the 20-hour window was moved backward by 3 minutes and 56 seconds each day. Second, in order to remove errors in antenna calibration, only one type of antenna, the Trimble Zephyr Geodetic with GP (TRM41249.00), was used for all static points as well as the CORS used to position the points. Lastly, a maximum of 40 kilometers from at least one CORS was viewed as a requirement for the best achievable accuracy. The Texas DOT operates a well distributed set of antennae that are part of the CORS network and are using the Trimble Zephyr Geodetic with GP (TRM41249.00) antennae. But a gap along the north end of the line existed. NGS installed two temporary CORS, one in Luling and one in Nixon, which allowed all points on the line to never exceed 40 kilometers from a CORS with the right antenna.
GPS also played a role in other parts of GSVS11. Originally, a real-time network (RTN) survey was planned, both to position the DIADEM camera offset marks and to provide some direct comparisons to LiDAR, imagery and the official 218 points. The Texas DOT initially provided NGS with equipment and access to its statewide RTN, but the loaned equipment had to be returned partway through the survey just as the RTN survey had begun, leaving NGS scrambling for an alternative. Because the primary purpose of the proposed RTN survey was positioning the DIADEM offset marks, the chosen alternative was to abandon the planned comparisons between LiDAR, imagery, static GPS and RTN. A static 20-minute GPS occupation was instead performed at each DIADEM offset mark, and OPUS-RS was used to position the point. New RTN equipment was located late in the summer, and the RTN component was eventually performed in October.
There were three components to the gravity survey as part of GSVS11. The first, and most urgent, were new absolute gravity values at each of the 218 marks. These were needed for reducing the leveling data and comparing against existing terrestrial data being used in gravimetric geoid models. A variety of methods was debated in the search for the right mix of absolute gravity measurements and relative gravity ties. . Originally, it was decided to take the NGS A-10 absolute gravimeter to each of the 218 points and forego relative gravity ties, but equipment failures led NGS to modify the plan and use primarily relative gravity, anchored to absolute gravity stations in Austin and Rockport. The complete A-10 survey on all 218 points was rescheduled for 2012.
The second component was to acquire vertical gravity gradients at each of the 218 points. These were needed both to reduce the A-10 measurements to the ground as well as to assist in reducing approximations about rock density and curvature of the plumb line for computing geoid models and reducing the DoV data to the geoid. A three-level platform system was adopted, using a LaCoste-Romberg G and D’s relative gravity meters in repeated measurements.
Finally, as a sort of gravimetric “anchor” on each end of the line, it was decided to establish a building-interior absolute gravity mark, measured with NGS’s FG-5 absolute gravimeter. The FG-5 point in Austin was downtown in the Stephen F. Austin building, which houses both the Texas General Land Office (TGLO) and the Texas Natural Resource Information System (TNRIS). The actual beginning of the survey line (the first of the 218 marks) was just outside this building. In the Corpus Christi area, the FG-5 mark was inside the Texas Maritime Museum in Rockport. The end point of the survey line (218 of 218) was a tidal bench mark just across the street at the Rockport tide gage.
However, once it was decide to take the FG-5 to Texas, two extra points were also chosen for measurement in Austin–Bowie High School and Bldg. 176 (the Institute for Geophysics “warehouse” building) of the Pickle Research Campus of the University of Texas. Pickle was chosen to build relations between NGS and the Bureau of Economic Geology, while Bowie was chosen because of an ongoing water-table experiment being conducted using a portable superconducting gravimeter onsite. For the sake of comparison, each of the four FG-5 marks were also surveyed with the A-10.
For the purposes of modeling the geoid and reducing the DoV data, an accurate DEM was needed along the survey line. Already in existence was the USGS 10-meter National Elevation Dataset (NED) data, but this was viewed as sub-optimal for the level of accuracy needed to establish this survey as a true calibration of the geoid. Instead, an airborne LiDAR-based DEM was proposed. Using the NOAA King Air aircraft and the Riegl Q680i-D LIDAR unit, flights covering the survey line were executed, with LiDAR coverage to about 0.5 kilometers on either side of the line and an approximate two returns per square meter post spacing (see Figure 7). The primary use of this data will be in the creation of an ellipsoid-height based DEM and a related orthometric-height DEM in the region of the survey.
A secondary use of this data will come from the flights in the Austin area. During one day of flights, GPS crews were scattered around Austin on 14 different predetermined bench marks. Those marks had been identified in two groups: Seven marks that were “good” (e.g., stable, first-order, having no evidence of vertical motion over time) and seven that were “bad” (e.g., unstable, lower order, having clear evidence of movement or possibly only a VERTCON-based NAVD 88 height). The purpose of this data collection was for NGS to study the various methods of using LiDAR for determining orthometric heights, including the use of GPS on vertical control (of various qualities) as well as no GPS on bench marks at all.
Although not needed explicitly for the reduction of any of the terrestrial geodetic surveys, digital airborne imagery was also collected over the survey.
This survey was established with a primary goal: To validate the accuracy of gravimetric geoid modeling at NGS. Although the delays in performing both the RTN survey and A-10 absolute gravity required temporarily deferring some of the planned studies, the number of datasets collected allowed for more than a dozen different additional investigations to be established. Of the studies remaining to be done are:
• Long-period GPS vs. OPUS-RS vs. RTN
• A-10 absolute gravity and FG-5 absolute gravity
• LiDAR-derived DEMs created in multiple ways
• LiDAR based positioning vs. RTN
• Photogrammetric positioning vs. RTN
• A-10 absolute gravity vs. relative gravity
Most of these are individual science investigations that will be published independently.
These studies have only just begun, and the results will not be available until sometime in late 2011 and through 2012. However, all of the data, having been collected by the federal government, will be made available to the public for use in additional studies. Furthermore, the legacy of GSVS11 is that the marks that were put in place are still there, and represent one of the most complete surveys in the nation’s history. It is NGS’s hope that these marks will have a permanence that allows them to be reused in additional surveys in the future.
Bibliography:Dracup, J. F. (1994). Geodetic Surveys in the United States, The Beginning and the Next One Hundred Years,1807 - 1940. Retrieved August 9, 2011, from National Geodetic Survey Web Site: www.ngs.noaa.gov/PUBS_LIB/geodetic_survey_1807.html
Heiskanen, W. A., & Moritz, H. (1987). Physical Geodesy (2 ed.). San Francisco: W.H. Freeman and Company.
Müller, A., Bürki, B., Kahle, H.-G., Hirt, C., & Marti, U. (2004). First Results from New High-Precision
Measurements of Deflections of the Vertical in Switzerland. In C. Jekeli, L. Bastos, & J. Fernandes (Ed.), Gravity, geoid and space missions: GGSM 2004, IAG International Symposium (pp. 143-148). Springer.
National Geodetic Survey. (2007). GRAV-D: Gravity for the Redefinition of the American Vertical Datum.
National Geodetic Survey. (2008). The National Geodetic Survey Ten-Year Plan: Mission, Vision and Strategy 2008-2018.
Saleh, J., Li, X., Wang, Y. M., Roman, D. R., & Smith, D. A. (2011). Error Analysis of the NGS Surface Gravity Database. Journal of Geodesy , Submitted for publication.
Sansò, F., & Rummel, R. (Eds.). (1997). Geodetic Boundary Value Problems in View of the Once Centimeter Geoid (Vols. Lecture Notes in Earth Sciences, Volume 65). Berlin: Springer.
Smith, D. A., & Milbert, D. G. (1999). The GEOID96 high-resolution geoid height model for the United States. Journal of Geodesy , 73, 219-236.
Stokes, G. G. (1849). On the variation of gravity on the surface of the earth. Transactions of the Cambridge Philosophical Society , 8, 672-695.
Project TimelineJuly 18, 2011 - Leveling and GPS work began
July 25 - Gravity work began
August 15 - DIADEM camera work began
August 8-12 - LiDAR and imagery flights took place
September 30 - All work completed, with the exception of leveling and gravity, which continued briefly into October, and absolute gravity, which was put off until early 2012