Photos by Patrick Koenig
By Derek Duncan
Bulldozers building golf courses have been common since the 1920s. Robotic bulldozers building golf courses are something new.
They’re being deployed in the highly anticipated construction of The Lido at Sand Valley in Wisconsin, and their technical capabilities mean, in theory, that the reproduction of the famed C.B. Macdonald course, which existed on Long Island from 1914 until it was demolished in the 1940s, can be executed with a measure of precision not previously considered.
When the Keiser family announced in early 2021 that they would be recreating The Lido in its exact (or as close to it as possible) form, historians and architecture buffs could have been excused for wondering exactly how that was going to be done. The confidence of building a golf course or almost anything else to a high degree of accuracy based off simplistic, 100-year-old plans that pre-date modern construction tools and techniques would be highly optimistic at best. And that’s if those detailed plans existed, which in the case of The Lido, they don’t. It also raises the philosophical riddle of whether a true replication, with no plans to measure it against and no person alive to testify to its veracity, could ever be verified.
As construction began on The Lido at Sand Valley in the fall of 2020, the prospect of accurate replication was dependent largely on a model of the course created by Peter Flory. Flory, an amateur architecture historian, spent years independently developing a virtual, 3-D computer simulation of the course based on all the historical data he could accumulate (photographs, aerials, written pieces). When brothers and Sand Valley proprietors Michael and Chris Keiser saw the finished product, they became convinced there was enough information in Flory’s richly rendered holes that recreating The Lido to scale, under the guidance of architect Tom Doak, was indeed possible.
Craig Haltom, golf course architect and president of Oliphant Golf, the Wisconsin-based company that’s provided construction support for each course at Sand Valley, including The Lido, agreed.
“Because [Flory’s model] was so photorealistic, and because of the way you could fly around in it in 3-D, I thought if we only had that we can always go out and use that as our reference,” Haltom says. “At the very least we’d have this visual representation of the golf course that would be useful.”
Still, some kind of foundation was needed to begin. In most cases the basic starting point for any golf course project is a topographical map. When the first Bill Coore and Ben Crenshaw course at Sand Valley was being planned in the mid-2010s, however, no topographical maps of the vast property were available.
So, Haltom reached out to Wisconsin Rapids-native Brian Zager, a software programmer who designed golf course simulations on the side. Zager had previously contacted Oliphant when he heard a golf resort was being developed near his home to see if his simulation-programming skills might be helpful—he regularly used publicly available LiDAR (Light Detection and Ranging) elevation scans when creating his computer-based golf courses. Perhaps that could be of use.
At Haltom’s request, Zager converted existing LiDAR imagery of the area into a physical topographical map of the property that Coore and Crenshaw would use for the design and construction of Sand Valley. He did the same at the David McLay Kidd-designed Mammoth Dunes, including building a simulation of the proposed land that was used as the template for Golf Digest’s Armchair Architect competition—the winning entry was made into the course’s current 14th hole.
For The Lido, the natural terrain of the property just north of the resort’s main entrance wasn’t very important—it would essentially be cleared of pine and levelled into the broad basin upon which Macdonald’s holes would then be raised. What Doak needed to build the course was a topographical map of The Lido’s actual features. Halton asked if Zager could produce a map not of the existing land (he’d already done that for the Keisers, previously) but of Flory’s model, which would provide the construction team with The Lido’s finished elevations at one-foot increments.
It’s one thing to use existing LiDar scans to create a contour map of actual land, but another to physically map a course that exists only on screen. The problem was that the two software programs—the one Flory used to create movable, multi-dimensional depictions of virtual courses, and the one Zager used that measured and triangulated millions of data points on a geological surface—weren’t compatible.
“The software [Flory] used was really cool, really high-end visually,” Zager says, but it wasn’t conducive to translating elevations. “It was quite difficult. There was a time when I didn’t think there was any way it was going to work.”
Zager continued to hack at the problem for nearly three months until, in September of 2020, he finally devised a way to capture digitally what the model was depicting. With a real contour map of the computer model’s simulated elevations in hand, Oliphant could stake the holes and proceed with rough grading the site so that Doak and associates Brian Schneider and Brian Slawnik would have a baseline to work off of when they arrived in the spring to do their shaping.
A MODEST SUGGESTION
As the rough work continued late into the fall and through the winter, Flory continued to make minor adjustments to the model. Zager then added the updates to his digital map, which were then printed onto paper. But because they were attempting to replicate another previously existent course down to each minute detail, the teams began to understand that more definition was needed.
Haltom asked Zager if he could convert The Lido’s previously established contours into a digital GPS canvas that could provide even more precise elevations for any given location on the property, effectively articulating the exact transitions between grades—the space from foot to foot. Again, Zager went to work. The program he devised effectively recalibrated the enormous accumulation of data points from his computer’s original scan of Flory’s model and converted them into a complete GPS map of the golf course that was accurate down to the inch.
Placing a mesh of minute coordinates across the entire site gave the bulldozer operators a verifiable sense of what each of their grades should be as they worked roughing in the holes, stopping frequently to cross-check elevations on the staked points before proceeding. At the same time, one day in late winter, an elderly man staying at a nearby resort stopped by to observe. He was a retiree who had owned a contracting company, and he asked why they were manually operating the bulldozers and checking stakes when they could use more efficient GPS-guided machines instead.
Contractors and civil engineers have been using GPS-guided bulldozers on various jobsites for years, but not until very recently have they been programmed to do the kind of feature-work on golf courses that artisan shapers have been doing for decades. The reason is that there first needs to be created a detailed illustration of the design—an intricately programmed set of instructions that’s accurate down to an inch or more—that can be translated into a language the machines can digest. In essence, a formula that can be read. In essence, what Flory and Zager had done. That’s rarely if ever been possible.
“The technology has existed for a while to build a golf hole plus or minus exactly to the plan,” Doak wrote to me in an email. “The problem is, and always has been, that nobody is able to draw a plan so well to capture all the little details that makes a hole come alive.
“Peter’s computer model was great—in most places it’s gotten us to within a foot of where we needed to be, and then we’d stare at the old pictures to hone it from there. All of his work saved me from trying to piece together the various elevations from the start.”
Once Zager had processed an accurate GPS survey of Flory’s model, they had what they needed to transmit the information to specialized bulldozers programed to receive it. Haltom contacted a local company to bring in a GPS-guided bulldozer for a demonstration on a section of what would become the 17th hole. They were quickly sold.
“What was surprising to us was the resolution you get from the machine at the first pass,” Haltom says. “It was putting in the small contours from Peter’s model that would typically never exist in a plan that wasn’t created in a 3-D environment. But because it was done in a video game, he was flying around and getting all the little details out of the GPS-programmed bulldozers.”
GPS-guided bulldozers still require an operator, who steers the machine across a section of land that will eventually become a green or bunker or fairway. But the program within the bulldozer reads the coordinates of its location on property and automatically adjusts the blade up or down, altering the grade with each pass until it’s refined to less than an inch of the intended elevation. When this writer visited in late summer, a large John Deere was shoving sand atop the massive Biarritz green, the eighth hole. The driver almost seemed bored.
As Doak mentions, however, there’s still a high degree of editing and refinement that he and Schneider and Slawnik need to make after the bulldozers produce the replica features of the model. Drainage, for instance—the way the water flows off of greens or collects in fairways—isn’t accounted for in the model, so trouble spots must be remediated.
“Almost every surface on that course has been ‘edited’ and re-graded by Brian and Brian and company to get it to a finished product, just like every other course I’ve ever built,” Doak says. In particular, the fairway contours didn’t match what the architects could discern in a 1926 aerial of The Lido and another from the 1930s, nor a photograph shot across the site at an oblique angle looking south revealing distinct shadow detail. These images also corresponded very closely to a clear photograph of a plasticine model Macdonald had made of the course, depicting similarly large fairway movements.
There is also the issue with green contours. While the aerials are invaluable at showing certain major features of the course—the size of the Biarritz or Punchbowl greens; the dimensions of the Channel hole; the turn angle of Raynor’s Prize Dogleg—they are less effective at interpreting what was going on with the putting surfaces. Flory’s model approximates slope and contour, but those approximations are based on a non-architect’s interpretation of grainy, two-dimensional black and white photos.
“The third dimension is the tough part,” Schneider says. “There was more work on the greens than we anticipated, including the outlines of the greens, where they sat, where the fall-offs were.”
He and Doak and Slawnik have spent decades extrapolating restorative information of historic courses from less-than-ideal photographs. At Sand Valley they’re relying not just on this kind of evidence for guidance, which is incomplete, but also on their accumulated knowledge of the greens of other Macdonald courses, specifically National Golf Links of America, his magnum opus where he developed his most significant architectural ideas just a few years before The Lido. And the faithful symmetry between what’s visible in the plasticine model and the fairway elevations Raynor actually built that show up in the aerials gives them additional confidence that what they can discern from the plasticine model’s greens contours are close to true.
About halfway through the build it occurred to the team that they could chart the data points of the elevations of the holes as they were being finished and make a virtual model of it that Zager could then upload back into the official GPS map of the course. Storing perfect information to the digital file would enable the bulldozers to replicate the post-production holes at any point in the future, even starting over from scratch if necessary. The potential to take a GPS snapshot of a fully graded and shaped hole, freezing it in time to be quickly reproduced in exact detail later illustrates the most optimistic use of GPS technology.
“[That] means, when I go to New Zealand in February [to begin work at Te Arai], I can shape all the greens on one visit and ‘save’ them, and I don’t have to worry if it’s several months later before they can get irrigation to the last green,” Doak says. “That will save me a couple of flights across the Pacific.”
The GPS bulldozers and subsequent modelling technology can provide efficiencies of exactitude and time. These, in turn, could lead to economic efficiencies, at least in terms of scanning a finished hole and preserving its precise contours, something that would naturally interest developers and clubs. Does it also present a way to design entire new golf courses on computers in the virtual realm, in which the simulated holes can be programmed into robotic bulldozers that build them perfectly as they advance across a piece of terrain? While he can see how GPS models might make paper topographic maps a thing of the past, Schneider doesn’t think we’re close to the day an architect could design an entire, complete course in three dimensions. At least not a great one.
“It would take a special kind of brain to be able to virtually imagine everything necessary to create something that special,” he says. “I don’t know if that kind of brain exists in the world just yet.”
The ability to adapt to the realities of a natural landscape, to be fluid and open to the contributions of team members—in other words, to not know everything that’s going to happen beforehand—is too fundamental to the way Doak, Coore and Crenshaw and other in-the-field designers build golf courses to change.
But what about replicating another entire course like The Lido, one that’s not extinct? Though it might not be practical, it’s theoretically possible to precisely map every elevation of an existing course, transform it into a GPS model the way Zager did and send that program to a fleet of GPS-guided bulldozers. An entire course could be recreated down to the inch if the inputs were correct.
In this future, would it be too far a stretch to imagine someone of extreme means finding a suitable site and building a 100-percent accurate replica of Augusta National? If Steve Wynn can construct 150-foot elevation changes lined with pine trees and cold-water streams on a flat desert site in Las Vegas—in 1989—couldn’t some Chinese billionaire cut into a sloping hillside near a river and do the same with a full-scale version of Augusta?
Doak thinks it’s far more likely the technology will enable architects to build specific features from elsewhere in exact form—a famous green or bunker—rather than an entire course, something he likens to sampling in pop music. But one way or another, the ability—and temptation—to clone either entire courses or particular features is fully developed. And that could trigger legal debates surrounding intellectual property. Do architects need to begin patenting certain greens and bunkers? Do famous clubs and courses need to copyright their golf designs?
Regardless of how the GPS mapping is applied elsewhere, it’s been adopted as part of the process of getting The Lido as close to the original as possible, a mission that the Keisers and the team have never let out of their sight.
“It’s been really inspiring how committed everyone has been to getting this right. When you’re walking across the golf holes it feels like St. Andrews out there,” Haltom says. “I’m a little surprised that we’ve been able to pull that off starting from [Flory’s] faithful reproduction and now building up to so many hours of bulldozers and trucks and real construction.”
He pauses for a moment.
“It’s amazing to think they built the original Lido it without all this heavy equipment.”
Photography provided by Patrick Koenig, check out his website here.