Coseismic vs. rainfall-triggered landslides: telling them apart in the field

A fresh scarp doesn't come with a label telling you what caused it. After a magnitude 6 event that hits during an already-wet season, you can end up with a slope population that has two, maybe three, different triggers tangled together. Sorting that out matters for the inventory you're compiling and for anyone downstream reading the hazard map, because the trigger changes what you'd expect to happen on that slope next time.

Spatial pattern is the first tell

Coseismic landslides cluster around the source in a way rainfall landslides usually don't. Keefer's classic threshold relations, and the follow-on work since, show landslide density falling off with distance from the epicenter and tracking shaking intensity more than it tracks rainfall totals. If you plot your inventory against isoseismal contours or a PGA raster and the density curve matches the attenuation pattern, that's a strong coseismic signature. Rainfall-triggered failures instead track storm totals and drainage structure. They crowd into specific catchments, convergent slopes, and areas downslope of saturated colluvium, regardless of distance from any epicenter. When you overlay both layers, shaking intensity and landslide density should correlate tightly for a coseismic event; for a rainfall event, that correlation falls apart and rainfall accumulation becomes the better predictor.

Slope aspect and geology also split differently depending on trigger. Coseismic failures show up on steep, often rock-dominated slopes with weak bedding or fracture planes oriented for sliding, the kind of setting where Newmark displacement analysis would flag low critical acceleration. Rainfall failures lean toward colluvium, residual soil, and slopes with a perched water table or a history of seepage. Neither rule is absolute. A slope primed by weeks of antecedent rainfall can still fail from shaking well below what Keefer's curves would predict, because the antecedent rainfall landslide trigger has already eaten into the safety margin before the ground even moves.

Scarp morphology and timing

Scarp freshness and shape help separate events that happened close together in time. Coseismic scarps tend to be angular, often following a joint or bedding plane, and they show up as a broad swarm appearing within the same satellite pass or the next clear one after the event. Rainfall-triggered scarps are more often curved, rotational failures with a toe bulge of debris runout, and they tend to appear in waves tied to storm cells rather than all at once.

The antecedent condition is the part that's easy to miss if you're only looking at the trigger event itself. A shaking event following a dry season behaves differently from the same magnitude event following weeks of rain. Pore pressure built up from prior storms lowers the shaking threshold needed to trigger failure, so you'll see landslides at shaking intensities where the dry-season Keefer curve says nothing should move. That's worth noting directly in the inventory record rather than assuming the trigger field is a clean either/or. Check the rainfall record for the 2 to 4 weeks before the earthquake, not just the day of.

Mixed trigger events

Mixed trigger landslide events are the ones that actually cause the most argument in a post-event survey meeting. An earthquake fractures and loosens a slope without producing an immediate failure, then the next rainstorm, sometimes weeks or months later, is what pushes it over. If your inventory only captures the scarps visible right after the earthquake, you'll undercount the slopes that are primed and waiting. A follow-up pass after the next significant rain event, compared against the earthquake-only inventory, is usually the only way to catch these delayed failures and tag them correctly rather than attributing them to whichever event happened to be most recent.

Practically, a regional inventory benefits from tagging each mapped scarp with the candidate trigger and the evidence for it, rather than forcing a single trigger label onto every record. Spatial correlation with shaking intensity, correlation with storm totals and catchment position, scarp morphology, and the antecedent rainfall window in the weeks before the event, taken together, usually point to one dominant trigger even in a mixed sequence. Scanning tile by tile for this kind of pattern across an entire affected region is slow going. Landslide Detection builds the region-wide scarp and debris-runout layer from high-res imagery after an event, so you're working from the full inventory rather than piecing it together valley by valley.

If you're compiling a post-event inventory and want the regional picture rather than a stack of single-site reports, get in touch about what Landslide Detection can map for your study area.