In 19th century America, the expansion of the railroad famously led to the creation of our modern time zones, imposing temporal order on a nation that to that point set time by dozens of local “high noons.” In bridging east and west and transforming America into a continental nation, the railroads also shaped the development of geodetic surveying. In the late 19th century and through the early decades of the 20th, multiple vertical datums — the standard for accurate measurement of elevation across large distances — existed across the country.

But over time these datums became consolidated. In a few years the key datum of the last 30 years, the North American Vertical Datum of 1988 — the datum used by federal agencies such as the Federal Emergency Management Agency and the U.S. Army Corps of Engineers — will be replaced by a new, satellite-based system.

To appreciate the ways in which the vertical datum ties the country together, it is helpful to know what it is. “A vertical datum is just a reference surface,” says David B. Zilkoski, the former director of the National Geodetic Survey at the National Oceanic and Atmospheric Administration and the man who helped supervise the monumental 14-year NAVD 88 project.

He uses the example of measuring someone’s height. You determine this by measuring from a person’s feet to the top of his or her head — the reference surface is the floor. “If you didn’t know that you’re measuring from where their feet touch the floor, you don’t have a good reference. So a reference is just where you start. You start somewhere and you say, ‘I know this number.’ In the case of measuring your height, you’re calling the floor zero.”

The vertical datum used by surveyors across the country is mean sea level, obtained through a tide gauge. Of course, measuring the sea is a lot trickier than measuring the floor in your house to check a person’s height. What sea are you leveling? Where? And how do you account for the fact that the sea, unlike the floor, is in constant motion? You cannot measure it. But you can model it by measuring data over 18.6 years to calculate a mean value. This accounts for sea-level changes caused by the sun, the moon, and local variations due to storms, he says.

So why do we need them? “Because when you’re going to build a road like an interstate ... if you are using a topographic map in North Carolina and then you cross into Tennessee and the maps are using different vertical reference frames, how would they match?” says Zilkoski. “You need a national datum for national projects. (For) anything that deals with height that is more than just local, you need to have consistency.”

According to a 1976 article by Ralph Moore Berry, then assistant to the director of the NGS, the first real geodetic leveling was conducted by the U.S. Coast Survey from 1856 to 1857, when a series of levels was run by G.B. Vose along the Hudson River between New York City and Greenbush (now Rensselaer), New York, across from Albany (“History of Geodetic Leveling in the United States,” Surveying and Mapping, June 1976, Vol. XXXVI, No. 2, pages 137-153). Berry described Vose’s confidence in his work: “From a hasty computation which I have made,” the surveyor claimed, “it appears that the probable error for the entire distance from New York to Greenbush does not exceed two-tenths of a foot.” Vose’s benchmark was used in 1875 to set the elevation of the Great Lakes.

Congress eventually authorized the U.S. Coast and Geodetic Survey (renamed in 1878) to lead a national geodetic project. The first benchmark in this surveying effort was placed at Hagerstown, Maryland, in 1877. “Using the Hagerstown benchmark as a starting point, benchmarks were set westward to assist in leveling the wild terrain,” wrote Aria Remondi, in one of a series of articles on NOAA’s website. “This enormous leveling undertaking supplemented the transcontinental arc of triangulation that followed the 39th parallel from Delaware across the country to California from 1877 to 1900. The data that surveyors gathered were extremely accurate given the tools they had and the conditions they endured.”

According to the NGS website, surveyors used a method called Differential or Spirit Leveling, where a rod is held on a mark with a known elevation. “A reading is taken on the rod, which provides the elevation of the line-of-sight through the instrument. Then the rod is moved to an unknown point and read again. Subtraction provides the elevation of the unknown point,” the site explains. Surveyors placed metal disks into the ground, benchmarks to delineate points of elevation across a territory. The work was exacting and tedious but essential.

Over the years, as more data came into the system from different locations, the C&GS conducted adjustments to ensure that leveling was based on a common reference. In 1900, for instance, the datum was adjusted when data from “C&GS, the U.S. Army Corps of Engineers, the U.S. Geological Survey, the Massachusetts Topographic Survey, and the Pennsylvania Railroad were all combined,” Remondi wrote. “Local mean sea level at five tide gauges was held fixed. Once all the data were based on the same datum, the organizations could reference each other’s data.”

Between benchmarks, placed about a mile apart, instruments were set up to measure the height difference between rods every 100 to 150 m. (Zilkoski explains that the datum’s official values are reported in metric units, though they are sometimes also provided in U.S. customary units.) A million or more observations were made in the continental United States, “and those elevation differences were combined in an adjustment to estimate one set of consistent values,” says Zilkoski.

Over time the amount of leveling increased. In adjustments conducted between 1900 and 1912, the amount of leveling the C&GS completed more than doubled, from 21,095 km to 46,468 km. The number of tide stations used as a reference also increased in the same span, from five to nine. “It was a lot of walking, a lot of hand notes; everything was checked twice,” says Zilkoski. He says old field reports would literally have two check marks. Data were collected, calculated, and archived by hand.

In 1929, the C&GS led a new datum adjustment, the Sea Level Datum of 1929 — an ambitious standard that would serve as the vertical reference point across the United States for more than 60 years. (The datum was later renamed the National Geodetic Vertical Datum of 1929.)

Fatefully, surveyors brought together 26 gauges — 21 in the United States and five in Canada — to calculate mean sea level. The tide stations formed a ring around the U.S. and Canada, encompassing points as far afield as Prince Rupert, British Columbia, and Biloxi, Mississippi. This greatly expanded the reach of the 1929 datum, but surveyors also knew that, as Remondi explained, “holding local mean sea level fixed at 26 locations would result in discrepancies since it was well known that mean sea level differed from location to location.” The decision was made on the assumption that variations “were probably about the same magnitude as observational errors, and that the fixed sea level would help avoid confusion amongst users if benchmarks near the coast did not agree with local mean sea level.”

But 625,000 km of new leveling data added since 1929 was still referencing the faulty 1929 datum. “We now know that significant errors were introduced into the 1929 General Adjustment by considering each of the tide stations to be on the same equipotential surface. The error is estimated to be as much as 0.7 m from coast to coast,” wrote Zilkoski and a colleague, Gary M. Young, in North American Vertical Datum (NAVD) Update.

The errors that had accumulated since 1929 ranged from refraction errors in observation (the shimmering effect on the surface of a road on a hot summer day can make it harder for surveyors to establish precise elevations) to post-glacial uplift in some states to subsidence from oil and water extraction in others. The inconsistencies of the 1929 datum piled up, and by the late 1960s, there was a growing chorus of surveying officials arguing that the 1929 datum needed a major adjustment. What’s more, many of the benchmarks that were part of the NGVD 29 network were destroyed during the building of the Interstate Highway System and the expansion of cities; Zilkoski says as many as 30 percent of those monuments were lost.

By 1973, the federal Office of Management and Budget concluded that the nation needed a new geodetic survey. Other agencies followed suit, and a new program began in fiscal year 1978 — the NAVD 88. The work of the NAVD 88 was handled by NOAA and Canada’s Survey and Mapping Branch.

The work, wrote Zilkoski and Young, required, for starters, that paper field records be converted to digital format with an acceptable error rate of less than 0.3 percent per data set. “Software was developed to read each data set and check for format errors or data omissions. Despite lost data shipments, a parcel delivery service strike, and tornado damage to the contractor’s plant, this contract was successfully completed in March 1977 with a minimum of data-set rejections,” wrote Zilkoski and Young. Archival leveling data, stored in 50,000 field books, also had to be converted for use on a computer.

From there, more than 80,000 km of the network had to be re-leveled. Destroyed survey monuments were replaced with stable “deep-rod” benchmarks. For the NAVD 88 — unlike the 1929 datum with its multiple tide gauges — Zilkoski and his colleagues picked a single tide station in Rimouski, Quebec, Canada (designated as Father Point/Rimouski), at the mouth of the St. Lawrence River. This decision enabled another datum, the International Great Lakes Datum, to share the same reference point.

“I didn’t want to constrain it to more than one tide gauge, because tide gauges don’t represent true mean sea level,” says Zilkoski. He performed studies to persuade his colleagues to base the datum on only one tide gauge. He believed that this would allow researchers to better model the remaining errors in the future.

The NAVD 88 used some of the original data from 1929 but also re-leveled about 100,000 km. Originally scheduled to be completed in 1988, it was finished on June 15, 1991. The new datum produced fewer distortions than earlier vertical datums and more accurate elevations, according to Remondi.

The NGS employed about 50 people in the field, who worked in six or seven crews to conduct the new leveling. Each crew required two rod people, an instrument person, a recorder/notetaker, and a supervisor. The work also required good shoes. “A friend of mine said he walked from Houston to Dallas because he leveled from Houston to Dallas,” Zilkoski says. “And at that time, you walked it.” In the West, surveyors didn’t let the Rockies or the Sierra Nevada stand in their way, traversing mountains by taking measurements at closer distances.

Despite the huge leap in accuracy from NGVD 29 to NAVD 88, the latter datum, like its predecessors, was still based on terrestrial survey marks that could be damaged or knocked out of position. According to the NGS, the 1988 datum was off by about 1 m coast to coast. It’s a small amount for such a large distance, but Zilkoski’s work marked the end of a long era. The new era will be defined by the Global Navigation Satellite System, which promises to produce orthometric heights more efficiently and accurately.

“I knew that GPS was going to replace leveling,” Zilkoski says. “I knew, and I kept working to try to make it happen.”

Remondi notes that the NGS began seriously looking into moving to GPS as early as 1994. According to a 2018 presentation delivered by NOAA regional geodetic advisers Bill Stone and Dana Caccamise, the benefits of switching include improved accuracy and consistency and easier operation. Basing the new datum on GNSS will reduce reliance on benchmarks on the ground. And it will still also leverage approximately 2,000 Continuously Operating Reference Stations —
remote, largely automated stations that can provide critical data about their exact positions and monitor and report slight changes.

Zilkoski sees NAVD 88 as the critical transition from one paradigm of surveying the world to another. “I always said that you needed to do the NAVD 88 if you’re ever going to use GNSS to obtain orthometric heights. You needed to get from NGVD 29, which had a lot of problems and issues, to a better vertical reference frame so that you could transition to the future based on GNSS.”

Speaking of the future, the NGS is developing transformation routines that will be used to transform NAVD 88 heights to the next vertical reference system. This will limit the need to visit all the benchmark sites in future datum adjustments. “Some of the benchmarks are in walls (and) buildings, and under trees,” says Zilkoski. “So, you can’t occupy them with a GNSS receiver. They don’t need them all for the transformation model, but they’re trying to occupy a benchmark every 10 km. And in some areas, every 2 km.”

Zilkoski retired from NOAA in 2009 and now runs a geospatial consulting firm. The conversion to the new system began in 2010 and is expected to wrap up in the next few years. In looking back on his long career, especially his time overseeing NAVD 88, he discussed his time in the field with the surveyors.

“You know,” he says, “I always enjoyed visiting the field personnel, but I’m not the type of guy to walk from Houston to Dallas.”

This article first appeared in the November 2020 issue of Civil Engineering.

Author