How to tell a looting pit from a natural sinkhole in satellite imagery

A fresh depression shows up on your archive comparison and the first question is always the same: is that a looting pit or did the ground just do what karst ground does. At 0.5 m resolution a hole is a hole until you read the context around it. Here's what actually separates the two, and where imagery alone runs out.

Shape, edges, and what's piled next to the hole

Natural dolines and sinkholes tend toward round or oval, with edges that slope gradually into the depression as the overlying material slumps. There's no spoil. The material that was there is still there, mostly, just lower and redistributed in a bowl shape. Weathering softens the rim over subsequent frames rather than sharpening it.

A looting pit reads differently. Someone dug it with a tool or a machine, so the plan shape tends toward rectangular, square, or an irregular polygon with straight-ish segments where a trench or test pit was cut. The edges are sharp in the first frame after the dig and often stay sharp for a season or two before erosion starts rounding them.

Most tellingly, there's backdirt. Excavated material gets piled beside the hole, not redistributed into it, so you're looking for a dark or light halo, depending on soil moisture and exposure, immediately adjacent to the depression. A doline doesn't produce a spoil heap. A looting pit almost always does, unless it's been swept or rained on hard enough to flatten the pile before the next pass.

Clustering matters too. Karst collapse follows the geology: dolines cluster along fracture lines, drainage paths, or wherever the epikarst is thinnest, and they show up more or less independent of what's buried nearby. Looting pits cluster on the archaeology. If the new depressions are strung along a tell's slope, sitting right on a known cemetery extent, or tracing the line of a buried wall visible in older frames as a soil-mark, that's disturbance following the site, not the rock.

Access tells you more than the hole itself

This is where vehicle tracks do a lot of the work that the pit shape alone can't. A natural sinkhole doesn't need anyone to drive to it. A looting pit, especially one large enough to be worth digging, usually does. Fresh tire tracks terminating at a new depression, a flattened path through vegetation, or a turnaround area nearby are strong secondary evidence, and they're one of the easier disturbance signatures to pick out at sub-meter resolution because they contrast against undisturbed ground in a way a single hole sometimes doesn't.

Timing is the other tell that karst can't fake. Sinkholes open up on their own schedule, tied to rainfall and subsurface drainage, and they don't reliably appear in tight clusters within a single review window unless there's been a specific triggering event like a flood. A set of new depressions that all appear between one quarter's pass and the next, concentrated on one part of a site, with access tracks leading to each one, is a pattern that karst processes don't produce on their own.

Where imagery stops and ground truthing starts

None of this is certain from the frame alone, and it shouldn't be treated as certain. Soil type, recent rainfall, and sun angle at capture can all make a natural slump look sharper-edged than it is, and a looting pit dug carefully with spoil scattered thin can look softer than you'd expect. Imagery narrows the list of depressions worth a closer look. It doesn't replace a site visit, a local partner's report, or a comparison against the known geology map for the area, and any flag that's going into a condition report to a funder or a national authority needs that ground truthing step before it's stated as fact rather than as an indication.

What imagery comparison is good for is making sure you're not scanning a full archive dump by eye, frame by frame, hunting for the one new pit or the one fresh set of tire tracks that wasn't there last quarter. Looting Monitoring runs that quarter-over-quarter comparison for a monitored site and returns the dated pairs where a pit or a track wasn't there before, so the review starts from a shortlist instead of a stack of raw passes. If you're tracking a site through a review cycle, that's the starting point worth a look.