Vehicle tracks in satellite imagery: normal traffic or looting access?

A site custodian walks a tell twice a week. A grazing family's truck cuts across the access road every market day. A looting crew backs a flatbed up to a fresh cut at 3 a.m. All three leave tracks, and from a single VHR frame they can look almost identical: two parallel lines of compacted or disturbed soil running toward the site boundary. What matters is what the track connects to, how often it repeats, and whether it's new.
Start with the baseline, not the frame in front of you
Most heritage sites in active conflict or post-conflict zones already have a traffic pattern, even unprotected ones. Staff vehicles, herders, farmers crossing a corner of the buffer zone, the occasional tour group before the drawdown. If you're looking at one image and trying to decide "is this normal," you're guessing. If you're comparing that image against the same ground from the prior quarter, you're not.
The question an officer actually needs answered isn't "are these tracks normal." It's "are these tracks new since the last review, and where do they lead." A track that's been there for three quarters, crosses the same spot it always has, and dead-ends at the custodian's hut is background noise. A track that didn't exist in Q2 and now runs straight to a disturbed patch near a known tumulus is the thing you flag.
That's a different task than eyeballing a raw image for "suspicious" vehicle activity, and it's the one that holds up when you write the incident note.
Reading the track, not just counting it
A few things separate routine access from an access route cut for looting:
Where it terminates. Tracks tied to staff housing, a guard post, a water point, or an established site entrance are administrative. Tracks that stop at a new depression, a mound with fresh spoil visible at its edge, or a section of wall that wasn't disturbed last quarter are not.
Whether it's paired with a pit. One new track by itself might be a surveyor, a patrol, a shepherd taking a shortcut. A new track paired with a new pit, especially one with a visible spoil heap, is a different signal entirely. The pairing is what you're looking for, not the track in isolation.
Repetition versus a single pass. A single thin line is one vehicle, once. Several overlapping, braided tracks in the same corridor usually mean repeated trips over days or weeks, which is consistent with a crew returning to work a site rather than someone passing through.
Direction relative to the perimeter. Tracks that run parallel to a known road or follow an existing field boundary read as civilian movement. Tracks that cut directly across undisturbed ground toward a specific feature, especially one that wasn't accessible by vehicle before, are harder to explain as routine.
None of these on their own are proof. Together, against a prior-quarter baseline, they're what turns a track from background into an entry in the incident log.
Why this is a quarter-over-quarter problem, not a single-image one
The honest answer to "how many vehicle tracks count as normal" is that there's no fixed number. Five tracks at a remote tell with no staff presence and five tracks at a site with a visitor center and a standing guard rotation mean completely different things. The method is constant across sites: pull the same ground from the last review cycle and check it against the current frame to see what changed.
That's the part that eats an officer's time when it's done by hand. Pulling the archived frame, lining it up against the current one, and walking the image corner to corner looking for a track that wasn't there before is slow, and it's easy to miss a thin new line against a cluttered frame, especially across a large monitored boundary with multiple access points.
Looting Monitoring is built around exactly this comparison: it checks VHR imagery of a monitored site across a quarter and returns a dated, flagged before/after pair showing fresh pits and vehicle tracks, so the baseline question (what's new since last time) has an answer before you open the raw frames yourself.
If you're tracking site condition across a conflict or post-conflict landscape and want that comparison run for you each quarter, that's the use case this is built for.