Reading the Mountain
I was browsing random wooded parcels in the point cloud viewer — the one I described in “The View from Every Angle” — when I noticed something weird.
At a certain angle — looking from below, oblique to the terrain — I could see faint linear paths running across the hillsides. Subtle enough to miss if you weren’t looking. My first thought was game trails. Deer and other animals do create detectable wear patterns over time, and I’d been curious whether LiDAR could resolve them.
I was wrong about what I was seeing. But figuring that out led somewhere more interesting.
What the Data Actually Shows
TALON integrates publicly available 3DEP LiDAR data from USGS for every parcel in its coverage area. This is airborne laser scanning — a plane flies over, fires millions of laser pulses at the ground, and records where they hit. Strip away everything except the returns that hit actual earth, and you have a bare-earth model of the land surface down to roughly one meter resolution.
Most land intelligence tools treat this as elevation data. Topography. Slope. That’s useful, but it’s the surface of what’s possible.
When I started looking at individual parcels in 3D — rotating the point cloud, changing viewing angles, filtering to ground points only — I started seeing things that don’t show up in any topographic map or county record. The terrain holds memory. Old disturbances leave marks that persist for decades, sometimes longer.
The features I was seeing weren’t game trails. They were too wide, too consistent, and they had a specific branching structure — multiple tracks converging at the bottom of each slope. That’s a logging road layout. A landing or drag point at the bottom, skid roads fanning uphill to reach timber.
Old-growth logging in Appalachia happened almost entirely this way. The operations stopped 80-100 years ago. The trees grew back. The roads didn’t disappear — they just got buried under a century of leaf litter and became invisible to everything except LiDAR.
Prototyping a Detector
Once I understood what I was looking at, I wanted to see if it was detectable algorithmically — not just visually.
The key insight is geometric. A skid road cut into a hillside is a bench: a narrow flat surface interrupting sloped terrain. The terrain goes steep → flat → steep in a span of 3-5 meters. That produces a detectable signature: sharp slope breaks on both sides of a smooth corridor. A drainage channel looks completely different — concave, channeled, water flows through it. A bench-cut road is flat, with edges.
I built a scoring pipeline that looks for exactly that pattern. For each cell in the terrain surface, it asks: is this cell locally flat? Does it sit in otherwise sloped terrain? Are there sharp slope breaks on both sides at road width? Does water accumulate here — if so, it’s probably a drainage, not a road. Are there other high-scoring cells nearby — because roads are linear, they don’t exist in isolation.
The result is a continuous score from 0 to 1 written back onto the original ground points as an extra dimension. Open it in CloudCompare, color by trail score, and the road network emerges.
Validation
The test was straightforward: run the detector on a parcel, pull up the winter aerial in Google Maps, and compare.
The correspondence was unambiguous. The high-score bands in the point cloud lined up spatially with the faint road textures visible in the aerial photography. More importantly, the point cloud resolved detail that the aerial couldn’t — roads that were barely visible as subtle texture differences in the imagery were clear, continuous linear features in the LiDAR output.
The parcel that started this whole thread — the one I thought had game trails — turned out to have an entire historical logging road network preserved in the terrain. Probably dating to the 1920s or 1930s based on the timber harvest history of that part of Appalachia.
Why It Matters
This isn’t integrated into TALON yet. What I built is a standalone research script — a prototype that proves the detection is possible with the data TALON already has.
But the value is real. Most buyers looking at wooded parcels in western NC, Virginia, or West Virginia have no idea whether historical access infrastructure exists on a property. County records don’t capture it. Sellers often don’t know. Aerial imagery hints at it in the best cases.
Old skid roads represent tangible value: reopening an existing compacted corridor costs a fraction of cutting new access, the network tells you something about the property’s timber history, and for hunting land buyers it defines where you can actually work the land.
TALON’s core purpose is surfacing information that doesn’t exist anywhere else in traditional real estate data. Historical access corridors fit that exactly. The data to find them has been publicly available for years — covering every wooded parcel in the country. It’s just been waiting for someone to look at it the right way.
That’s what we’re building.