Debris flow vs rockfall: how to tell them apart in satellite imagery

After a storm or a shaking event, the imagery comes in and the inventory has to get built fast, usually before the next rain event wipes out the freshest signatures. The first sorting problem is almost always the same one: is that scar a debris flow or a rockfall. They often sit on the same slope, sometimes triggered by the same storm cell, and at 0.5-2 m resolution the two can look similar enough to cost you a field day you don't have.

Shape and runout tell the story first

A rockfall runout is blocky and fans out in a cone or lobe below a cliff or steep rock face, usually below a clean, angular headwall scarp. The debris pile sits close to the source, the runout distance is short relative to the drop height, and individual boulders are often resolvable as discrete bright or shadowed clasts scattered across the surface, especially at 0.5 m. There's rarely much sorting, so large blocks and fine material end up side by side right at the toe.

A debris flow behaves differently because it's water-charged and it travels. Look for a channelized track: a narrow, often sinuous scar that follows an existing drainage or gully, with levees visible as parallel ridges along the margins where the flow stripped vegetation on its way down. That channelized debris flow signature is the single most reliable discriminator in optical imagery. Rockfall doesn't follow drainages, because it doesn't need water to move. Debris flows almost always do.

Color and texture add a second layer of confirmation. Fresh rockfall exposes light-colored, unweathered rock on the headwall, often a sharp contrast against darker surrounding bedrock or soil. Debris flow tracks tend to read as a muddy brown or gray stripe cutting through green vegetation, because the flow scours a corridor rather than exposing a rock face.

Talus cone vs debris fan at the toe

The depositional landform at the bottom is where people second-guess themselves most, especially a season after the event, once the sharpest tonal contrast has faded a bit.

A talus cone builds from repeated small rockfalls accumulating at the base of the same cliff over years, so it has a symmetric, radial shape centered directly below the source and a fairly uniform, coarse texture. A fresh event on top of an old talus deposit shows up as a tonal change confined to the upper portion of the cone, right where the new material sits.

A debris fan is what a channelized debris flow builds where the gully opens onto a valley floor or a fan apex. It spreads laterally in a broader arc, often with a visible apex where the channel widens, and the surface can show braided or lobate texture from multiple flow pulses instead of one uniform accumulation. If the fan sits downstream of a narrow drainage with levees further up the track, that's a debris flow fan. The drainage connection upstream is usually the detail that settles the borderline calls.

Where this breaks down

Mixed events happen. A rockfall can dump enough loose material into a channel carrying water to trigger a debris flow downstream, so you'll sometimes see a blocky rockfall scar feeding directly into a channelized track lower on the slope. Code that as a composite event rather than forcing it into one category. For a regional inventory, flagging the mixed cases separately usually matters more than getting a clean single label on every scar, because they point to a different downstream hazard than either process alone.

Shadow geometry and sun angle distort apparent runout shape too. A scarp that reads short and blocky in one scene can stretch out more in a follow-up acquisition with a different illumination angle. Cross-checking two dates, where the archive allows it, catches most of these false calls before they go into the inventory.

None of this replaces a field visit for the sites that matter most, but for a region-wide first pass after a storm or an earthquake, getting the debris flow and rockfall calls right from imagery alone is what lets a survey team decide which valleys actually need boots on the ground. Landslide Detection builds that regional first-pass layer from high-res satellite imagery, so the inventory starts with the whole region mapped rather than one valley scanned at a time.

If your team is compiling a post-event inventory this season, it's worth seeing what a region-wide scarp and runout layer looks like before anyone starts scanning archive tiles by hand.