The Ribbon That Refused to Touch Its Own Terrain

Dear version of me who hasn’t tried this yet,

My hobby is collecting hobbies, and hobby number eleven is VFR Track Relief Printing — exporting a GPS log from a flight, merging it with elevation data, and generating a 3D-printable relief tile where the terrain rises and the track is engraved like a thin ribbon across the landscape. You’re going to spend the first hour convinced you’ve figured something out. GPS log from last month’s Wetaskiwin run, elevation data from a NASA server, a Python script cobbled together during lunch. The preview on screen will look perfect — Cooking Lake’s shoreline rising gently from the plain, the flight track threading between terrain you remember avoiding.

Then the print will fail. Three times.

What nobody tells you about merging GPS tracks with elevation models is that they disagree about where the ground actually is. GPS altitude references the WGS84 ellipsoid — an idealized mathematical surface — while DEM data like NASA’s SRTM measures orthometric height from mean sea level. The difference between these two reference frames is called geoid undulation, and in central Alberta it runs about 21 metres. Your track will float above its own terrain like a ghost unless you apply a correction, and the first time you see your careful ribbon hovering a storey above the ridge you distinctly remember clearing by fifty feet, you’ll spend an hour checking your coordinate parsing before discovering this is a geodesy problem, not a bug.

There’s a strange echo here from the orthomosaic mapping work — both hobbies turn flight data into spatial documents you can argue with. But where the foam wing captures terrain from above and reconstructs it from parallax, this reverses the flow: terrain data already exists, and what you’re adding is the thin line of a single human decision. Where did I fly? What did I avoid? The tile becomes a record of judgment calls.

The SRTM data itself has peculiarities worth knowing. The Shuttle Radar Topography Mission flew in February 2000 — a single eleven-day mission that mapped 80% of Earth’s landmass using synthetic aperture radar. But radar casts shadows. Steep slopes facing away from the shuttle’s orbit angle appear as voids or spikes in the data. Those artifacts land exactly where interesting terrain tends to be: the north faces of foothills, the shadowed side of river valleys. Your first tile will probably include at least one anomalous spike where the terrain data hiccuped, and you’ll have to decide whether to smooth it out or leave the scar.

Then there’s vertical exaggeration — a problem I hadn’t anticipated. Every physical relief map you’ve ever touched exaggerates the vertical axis. Typically two to ten times horizontal scale. Without it, terrain looks flat because it mostly is; Alberta’s foothills compress to a barely perceptible wrinkle at 1:1. But exaggerate too aggressively and your track embeds into ridges it actually cleared by comfortable margins. The ratio is surprisingly subjective. I’ve printed three versions of the same flight at 2×, 5×, and 8× exaggeration, and none of them capture exactly what I remember seeing through the windscreen.

What I wasn’t expecting: how much orientation matters during printing. The flight track groove is shallow — maybe 0.8 mm deep — and the printer builds in layers 0.2 mm thick. If the groove runs perpendicular to the layer lines, each layer deposits cleanly into it. Rotate the model 45 degrees and those same layers create staircases that blur the path. The longest axis of the terrain should align with the print bed’s X-axis, and even then some passes of the track will print cleaner than others depending on their angle.

A Swiss engineer named Franz Ludwig Pfyffer spent 24 years in the eighteenth century carving a 1:11,500 relief model of central Switzerland from wood and plaster. It took him until 1786 to finish. I got a rough equivalent in eleven hours of compute and five hours of print time, which probably says something about the compression of history, though I’m not sure what.

The tile sits on my desk now, white PLA, slightly warped at one corner where the bed adhesion slipped during hour three. The track is there — a thin vein through terrain that I recognize, mostly. Cooking Lake catches the light differently than the surrounding plain. The foothills to the west rise just enough to cast tiny shadows. When I run my thumb across the surface, I can trace the exact moment I turned crosswind at Wetaskiwin, the shallow descent into Cooking Lake’s circuit, the final approach where the runway appeared exactly where it should have been.

It’s not quite what I saw from the air. The compression is wrong, the exaggeration is debatable, and the geoid correction might be off by a few metres. But it’s the flight, held in plastic, and that’s more than I had yesterday.