Earthquake-induced landscape preconditioning from a 6-year multi-temporal analysis of the 2018 Mw 7.5 Porgera earthquake region, Papua New Guinea

Earthquakes pose significant threats to mountainous regions, where co-seismic ground shaking and topographic amplification along ridges can trigger hundreds to thousands of landslides, a damaging and widespread impact with implications for relief and reconstruction efforts. While the spatial distribution of co-seismic landslides has been extensively documented from numerous worldwide studies, questions remain regarding long-term landscape preconditioning caused by large earthquakes. Preconditioning occurs when seismic events cause elevated landslide rates above the normal background rate that can persist for up to a decade after the initial trigger. Remote sensing methods utilizing optical satellite imagery enable the development of multi-temporal inventories that characterize slope failures across pre- and post-seismic periods, revealing how peak ground acceleration (PGA) combined with excess topography (landscape zones above a stable threshold slope) can precondition landscapes for future instability. Papua New Guinea (PNG) experiences frequent large earthquakes, with concomitant landsliding and other environmental effects, for example in 2018 the Mw 7.5 Porgera earthquake was reported to have generated ~11,000 co-seismic landslides. This study investigates the sustained effects of PGA on the landscape evolution of PNG. Using false colour composites derived from band-ratio manipulation of high-resolution PlanetLabs imagery, combined with control factors, manual mapping was conducted using systematic visual comparison of pre- and post-event imagery.  A new 6-year multi-temporal inventory for PNG is presented, documenting over 6,000 landslides across pre- and post-seismic periods, with a landslide average of ~700/yr pre-earthquake and a maximum of 347 per year post-earthquake. The majority of slope failures occurred during the immediate monsoon season following the 2018 earthquake, exceeding 4,000 events. Analysis of triggering factors revealed non-linear relationships with rainfall and strong negative exponential correlations with stream distance, identifying variables with a greater influence upon landslide susceptibility. Probability-density analysis displayed low rollover thresholds and narrow quantile bands, indicating high inventory completeness and consistent data distribution. Moreover, consistent with previous studies, the landscape exhibits a rapid recovery period, demonstrating short-term preconditioning, with landscape disturbance persisting for only one year before returning to pre-seismic conditions. This multi-temporal dataset for PNG provides significant insights into landslide distribution patterns, enhancing our capacity to forecast post-seismic landslide activity, contributing to more robust susceptibility assessment frameworks for seismically active mountainous regions.