Contrasting slope histories under similar final-window rainfall in rock-slope failures: implications for monitoring

Rainfall is central to slope-hazard assessment and early warning, yet the meteorological state at release may not represent the pre-failure history of a rock slope. We reconstructed failure time, rolling precipitation windows, antecedent forcing, and precursor evidence for a purposively selected set of 12 natural rock-slope failures with resolvable event timing and meteorological evidence. Six events provided quantitative event-window rainfall: five had P24 = 0 mm and Pengshui had P24 = 0.9 mm. A sensitivity check across 12-, 24-, 48-, and 72-h windows showed that earlier rainfall re-entered the longer windows for some events, confirming that P24 is a standardized immediate descriptor rather than a trigger threshold. Within the six quantitative cases, antecedent histories ranged from multi-week dryness to earlier extreme rainfall, reservoir-level cycling, and short-lag rainfall, while documented precursor evidence ranged from multi-week deformation to minute-scale block shedding. Six additional boundary cases separated physical-process boundaries from evidence-definition boundaries. Because the search and retention criteria intentionally favored events with resolvable timing and meteorology, the set is not representative and is not used to estimate prevalence. The event-level comparison instead shows that a weak or absent final rainfall signal can coexist with contrasting pre-failure histories and monitoring evidence.