Temporal Variations in Landslide Distributions Following Extreme Events: Implications for Landslide Susceptibility Modeling

Landslide susceptibility models are fundamental components of landslide risk management strategies. These typically assume that occurrence is time-independent, even though processes including earthquake preconditioning and path dependency transiently impact occurrence. Understanding the temporal characteristics remains limited by a lack systematic investigation into how distributions vary through time, this impacts susceptibility. Here, we apply Kolmogorov-Smirnoff chi-square statistics to 30-yr inventory monsoon-triggered landslides from Nepal systematically quantify spatial time in “normal” years impacted extreme events. We then develop binary logistic regression (BLR) for 12 yrs our with >400 use area under receiver operator curve validation assess well these can hindcast other years. found particularly storms (1993 2002), earthquakes (2015), floods (2017). Notably, Gorkha landscape shifted 2015 higher slopes, reliefs, excess topographies. variations significantly BLR modeling, trained on unable consistently However, developing using increasingly long historical inventories shows developed >6–8 data provide good hindcasting accuracy. Overall, results challenge time-independent assumptions approaches, highlighting need time-dependent modeling techniques or modeling.