A diver “flies” with an electric propulsion vehicle known as the “Fat Man” at the 80-foot level of Wakulla Springs. A transfer capsule (behind) awaits his return.

Located in Florida’s panhandle, Wakulla Springs is a mysterious place. An underground river suddenly materializes deep in the heart of a cypress forest. Alligators, “snake birds” and pre-historic looking fish call it home. It’s where Hollywood shot Creature from the Black Lagoon and the original Tarzan movies. It’s also the site where one of the world’s most extreme scientific and exploration-related diving projects took place—through the river’s mouth and into a huge underwater cavern known as the “Grand Canyon Dome.” Today, the underground cavern and its labyrinth of tunnels is part of the world’s first three-dimensional digital map of an underwater cave, created during the dive project using cutting-edge technology, including a 3D Digital Wall Mapper (DWM) and the Global Positioning System (GPS).

It was in this remarkable area that an amazing project took place—the three-month Wakulla 2 Expedition of the National Geographic Society, which began in December 1998. I was the co-leader of the expedition, which attracted some of the world’s best divers and engineers, and which used advanced technology to create the first 3D underwater cave map.

For years no one knew where the river came from, nor what really lay under the swampy surface. Divers first explored the underground river in the 1950s and by 1987 only 250 meters of the underwater labyrinth were known. In 1987, the National Geographic Society sponsored a 10-week project to explore and map the unknown reaches; 3.3 kilometers of new tunnels were discovered and tentatively mapped. But mapping techniques prior to the Wakulla 2 project were often crude and imprecise—line-and-compass—and weren’t connected to any surface features, let alone registered to any United States surveying and mapping datum for precision location.

The Wakulla 2 Project was funded by private donations and NGS to the tune of nearly $1 million, most of which financed the DWM, developed specifically for the project. The project’s main goal was to create the 3D virtual reality map. This interactive “swim through” would be used to better understand the underground caverns, how water travels through them, and how to preserve them in the face of inevitable surface development.

Setting the Mapping Stage

The Wakulla project was a technological masterpiece, employing some of the most advanced technologies ever used in mapping an underwater cave system. More than 150 people, including some of most experienced cave divers, scientists and engineers from all over the world converged on Wakulla, seeking to unveil its mysteries. All but one were volunteers.

“It’s the first time anyone tried to do a complex precision map under these conditions,” says Brian Pease, developer of a radiolocation system used to precisely locate cave positions on the earth’s surface. “It’s a one-entrance cave, not like those in the Yucatan that have multiple openings. Wakulla goes down deep and stays deep. The challenge was in the great length, depth and huge size of passages, all served by a single entrance.”

Cave Radios, GPS and Other Technologies

Pease was to provide what is known in caving circles as radiolocations, points located on the earth’s surface exactly above low frequency (3500 Hz) magnetic induction “beacons” placed in the cave spring by divers. The estimated accuracy of the radiolocations vary from 0.1 m to 1 m, the lower accuracy occurring where there is high power line interference. Once these points are precisely located on the surface, they provide known positions in the passages 90 m below.

Because of the precision required, Pease needed a surface location system that would be rugged, accurate (to cm level) and simple enough to use in the swamp and heavy brush terrain of Wakulla Springs. After researching products, he selected Trimble’s (Sunnyvale, Calif.) GPS Total Station 4800 survey system.

“Conventional surveys depend on every point before the one surveyed,” Pease says. “Using a conventional system in this terrain would have been a nightmare. The cave spread out over five square kilometers; it would’ve taken a long time if we had used conventional surveying. We just didn’t have the time and manpower to clear brush from several kilometers of sight lines.”

For the underwater radiolocation beacon, Pease used a 6-watt 3.5 kHz oscillator connected to a tuned induction coil antenna. Overhead, Pease carried a phase-locked loop directional antenna to pick up the electromagnetic signals through the rock, which could be detected up to 600 meters horizontally from the beacons. It was a relatively simple system with a profound outcome when combined with the DWM and GPS.

In some ways the whole thing was as spectacular as if it were black magic. The magnetic induction transmitter pulsed through 600 meters of rock; we surface-fixed with GPS and nailed the world coordinates. Before, we didn’t know where the cave tunnels were. With GPS, we did.

Cave Diving

Cave divers represent an extremely small fraternity, requiring extreme skill, discipline and courage. Less than one percent of all divers qualify for cave diving. Within the cave diving community, Wakulla Springs is considered to be one of the great remaining unexplored frontiers. Distances in excess of 4 km from the entrance at depths of up to 100 m underwater have also made it one of the most formidable of challenges. World-class cave divers came from as far as Australia, Germany, Denmark and the Bahamas to join the team.

The dive equipment was cutting edge, using some of the most advanced technology ever used in diving. Rebreathers precisely regulated the amount of oxygen in the breathing gas and enabled longer dive times. A decompression chamber floated above the cave entrance on a barge; after each dive, a transfer capsule pressurized to the divers’ depth would be lowered to carry them to the decompression chamber above the surface.

Divers remained underwater on average of three to four hours at a bottom depth of 84 to 91 m (275 to 300 ft). Recovery time included up to 13 hours in the chamber. The average total dive time from surface to decompression was 20 hours round trip. In total, the team logged 234 rebreather missions for a total of 661 hours on rebreathers, including mapping and radiolocation dives as well as training and dress rehearsal dives. More than 500 open-circuit diving missions also supported the rebreather missions.

Placing the Beacons

The Wakulla dive project was precisely planned and carefully practiced in the Wakulla Springs basin prior to the actual dive into the hidden caverns. Each mission mapped new reaches of the tunnels, some of which stretched more than 3,000 m. Each section needed to be traversed at least twice; once to place the radio beacon, then to “map” and locate the points and pick up additional visual data.

Each mission was time-intensive; placing the beacons took up to five hours, with an average of 10 hours for decompression. Usually only two beacons were placed; the longest mission placed four. To enable the beacons to work underwater, Pease mounted each loop antenna on a hollow plastic frame, providing rigidity, floatation and protection for the copper wire inside. It also contained a bubble level, used to level the loop. The antenna was attached to a transparent housing that contained the 3.5 kHz oscillator and a battery.

Usually, the divers would place the beacons at pre-determined locations, switch them on and swim away. Most of the beacons were set to go off in five minutes. Pease had to be in the general area overhead with the antenna to pick up signals. Other beacons were set to activate hours later, usually the next morning when atmospheric interference was lowest.

On the next dive, or “fly-by,” divers would tag the data by hitting a trigger on the DWM when they flew directly over the beacon. To coordinate with surface locations, divers also noted the number attached to each beacon. When locations seemed imprecise and suspect, divers redid the point.

Brian Pease establishes radio fixes to ensure accuracy of the 3D digital map.

The DWM in Action

The DWM is a complex instrument controlled by eight onboard computers. It uses sonar technology and the same inertial navigation device (IMU) used by commercial fighter aircraft for missile guidance. The DWM is mounted on an underwater scooter that propels divers at up to 1.3 knots and for distances of up to 20 km on one battery. Lead divers drove the DWM and carried back-up propulsion units.

The DWM emits as many as 32 simultaneous sonar pulses in a circular (spiral) pattern, at a rate of up to four times per second. An onboard computer keeps track of how far the sonar travels, adjusting for the pitch and roll of the scooter. The data provided the exact dimensions of the cave tunnels, including walls, ceiling and floor.

“Traditionally, divers are lucky if they get any data for cave walls. Period,” says Barbara am Ende, Ph.D., mapping expert for the project. “The sonar picked up every nook and cranny.”

When it came to precise positioning of the data, however, the IMU poised a potential problem. Best suited for an aircraft’s high velocity, the IMU had too great a drift buildup at the slow dive speed; the resulting inexact data needed to be corrected for the x-y drift. That’s where the radiolocation beacons came in.

By using the beacons as surface-fixed control points, they were able to register the 3D digital map to these high precision waypoints. And to further reduce drift buildup, they developed a way to acquire data segmentally from waypoint to waypoint. After two to three of the segments in each mission, the divers would drop the DWM to the cave floor, its position coinciding with the radio beacon location and reset its coordinates to [0,0, depth]. Am Ende would later replace the [0,0] values for the beginning of the segment with the GPS-fixed coordinates for the point. By comparing the IMU-derived coordinates for the end of the segment with the GPS-fixed coordinates, the mapped data in between could be adjusted to compensate for the IMU errors. This method for compensating for the drift of the IMU turned out to be highly effective.

Am Ende used software developed primarily by team member Fred Wefer to process the DWM tunnel scans to create the 3D cave portrayal, which included the cave’s depth, width and contours. Each day, am Ende would discuss the data with me and Pease to determine where they needed additional radiolocation points to fill in the data. Every major passage junction was a critical area; other areas included bends, turns and any points appearing imprecise on first mapping. They plotted more than 40 radiolocation points and nearly 10 million wall, ceiling and floor points.

Alligators and Other Obstacles

A critical step of the mapping project was in determining the beacon’s positions on the surface in real world coordinates. To attain the high accuracy (cm level) Real-Time Kinematic (RTK) positioning of these points, horizontal control was extended from a high accuracy reference network (HARN) point 10 miles away, using static GPS observation techniques and post-processing. The control was extended to a reference mark in the concrete floor of the dive platform located over the cave entrance at Wakulla Springs. A 4800 GPS receiver placed on a tripod over this mark was then used as the RTK reference receiver. The corrections generated by the reference receiver were sent to the roving RTK receiver via a 25-watt UHF transmitter that had its antenna on a 20-foot pole at the top of the dive platform. These correction signals could be received anywhere in the park.

Radiolocated points were both inside and outside the park. Each area had its challenges: inside the park the forest floor was clear and free of most underbrush, but the huge, old-growth cypress trees caused problems at times with GPS reception, especially the L2 frequency. When this happened, Pease would locate pairs of reference points in nearby open areas using the GPS receiver as a base. I later used an optical total station to finish the survey. Outside the park, the land had been logged and the sky was clearer, but Pease often had to clear away underbrush and chop down small trees at the point’s location.

“I would charge through the underbrush, chopping at everything in my way with my machete, looking for the location with the directional antenna,” Pease says. Once there, he planted a pole and later used RTK GPS to fix its real world position.

“It saved us an enormous amount of time and labor,” he says. “Because it was RTK, we got the data right then and knew it was right. When we left the park, we took most of the marks with us, so we had to get solid data while we were there.”

One point was in the middle of a river. Pease used a canoe to get there, then jammed a pole into the bottom to locate the beacon’s signal. Given the circumstances, foot-level accuracy was satisfactory.

For another point on the far side of the river, Pease had to pole his canoe through the shallow waters and mud. If they had used the tripod to support the GPS receiver, the battery would have been underwater. Instead, they set the receiver on a tree limb so that the GPS antenna would be exactly over the radiolocation point. Pease operated the receiver—and his partner watched for alligators.

“The real beauty is there is no cumulative error,” Pease says. “Every point is going to be exactly right; you’re not depending on any other point. If something comes out bad, you know it right away. It’s really what makes GPS so good. It literally works in the middle of nowhere.”

Mapping the Mystery

Am Ende headed up the computer mapping. In all, the project took 10 months to complete. Getting the waypoints was one thing; creating a precise, seamless map was another. And GPS was critical for the accuracy of the outcome.

“Because of the GPS-fixed radio locations, we could tell the computer that this sub-surface point is exactly below this surveyed point on the surface,” am Ende says. And since the depth gauges were reliable, they knew exactly where those points were in the cave.

“What I had to do was force the IMU-derived path of the line of travel determined by the DWM to go through that exact point,” she says.

For example, if she knew exactly where points 11 and 12 lay (i.e., radiolocated and GPS-fixed), and had at least two mapping paths that included those points (ingoing and outgoing), she could overlay the lines and see the difference due to drift.

“I would decide whether one looked more right than the other,” she says. “Or I would split the difference in between and tell the computer where the point would be.”

In practical terms that means if you’re walking above the cave, each GPS-located radiolocation point would show you precisely what lies under the surface. In between, however, while the measurements of the cave’s shape are good, the surface locations are just a good guess. “We’re not sure exactly except for right under the GPS points,” she says.

“Before this project I didn’t know much about GPS,” am Ende says. “But I can’t overstate how important it was to the success of the project. We wouldn’t have what we have without it. There was no other way of correcting errors due to inertial drift of the DWM navigation systems. It was our way of correcting to reality, connecting what was going on in the cave to a real world coordinate.”

Ten months of post-processing rendered this full technicolor map of Wakulla Basin in 3D.

So what’s next?

“Our goal now is to mesh the points into polygons to create a solid surface for a virtual fly-through the cave,” am Ende says. The fly-through should look substantially like being in the cave itself, as am Ende will “drape” photo images taken during the project onto the “mesh walls.” Future visitors to Wakulla Springs will be able to slip on a helmet and goggles and take their own trip through the Grand Canyon Dome.

“It won’t be exact,” she says, “but it’ll come close.”

William C. Stone, Ph.D., P.E., is the leader of the Construction Metrology and Automation Group at the National Institute of Standards and Technology (NIST) in Gaithersburg, Md. He is also president of the U.S. Deep Caving Team and served as co-leader of the Wakulla 2 Expedition.