The addition of water on or below the earth’s surface generates changes in stress that can trigger both stable and unstable sliding of landslides and faults. While these sliding behaviours are well-described by commonly used mechanical models developed from laboratory testing (e.g., critical-state soil mechanics and rate-and-state friction), less is known about the field-scale environmental conditions or kinematic behaviours that occur during the transition from stable to unstable sliding. Here we use radar interferometry (InSAR) and a simple 1D hydrological model to characterize 8 years of stable sliding of the Mud Creek landslide, California, USA, prior to its rapid acceleration and catastrophic failure on May 20, 2017. Our results suggest a large increase in pore-fluid pressure occurred during a shift from historic drought to record rainfall that triggered a large increase in velocity and drove slip localization, overcoming the stabilizing mechanisms that had previously inhibited landslide acceleration. Given the predicted increase in precipitation extremes with a warming climate, we expect it to become more common for landslides to transition from stable to unstable motion, and therefore a better assessment of this destabilization process is required to prevent loss of life and infrastructure.

Catastrophic landslides cause billions of dollars in damages and claim thousands of lives annually, whereas slow-moving landslides with negligible inertia dominate sediment transport on many weathered hillslopes. Surprisingly, both failure modes are displayed by nearby landslides (and individual landslides in different years) subjected to almost identical environmental conditions. Such observations have motivated the search for mechanisms that can cause slow-moving landslides to transition via runaway acceleration to catastrophic failure. A similarly diverse range of sliding behavior, including earthquakes and slow-slip events, occurs along tectonic faults. Our understanding of these phenomena has benefitted from mechanical treatments that rely upon key ingredients that are notably absent from previous landslide descriptions. Here, we describe landslide motion using a rate- and state-dependent frictional model that incorporates a nonlocal stress balance to account for the elastic response to gradients in slip. Our idealized, one-dimensional model reproduces both the displacement patterns observed in slow-moving landslides and the acceleration toward failure exhibited by catastrophic events. Catastrophic failure occurs only when the slip surface is characterized by rate-weakening friction and its lateral dimensions exceed a critical nucleation length <mml:math><mml:mrow><mml:msup><mml:mi>h</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math>h* that is shorter for higher effective stresses. However, landslides that are extensive enough to fall within this regime can nevertheless slide slowly for months or years before catastrophic failure. Our results suggest that the diversity of slip behavior observed during landslides can be described with a single model adapted from standard fault mechanics treatments.

Episodically to continuously active slow‐moving landslides are driven by precipitation. Climate change, which is altering both the frequency and magnitude of precipitation worldwide, is therefore predicted to have a major impact on landslides. Here we examine the behavior of hundreds of slow‐moving landslides in northern California in response to large changes in annual precipitation that occurred between 2016 and 2018. We quantify the landslide displacement using repeat‐pass radar interferometry and pixel offset tracking techniques on a novel data set from the airborne NASA/JPL Uninhabited Aerial Vehicle Synthetic Aperture Radar. We found that 312 landslides were moving due to extreme rainfall during 2017, compared to 119 during 2016, which was the final year of a historic multiyear drought. However, with a return to below to average rainfall in 2018, only 146 landslides remained in motion. The increased number of landslides during 2017 was primarily accommodated by landslides that were smaller than the landslides that remained active between 2016 and 2018. Furthermore, by examining a subset of 51 landslides, we found that 49 had increased velocities during 2017 when compared to 2016. Our results show that slow‐moving landslides are sensitive to large changes in annual precipitation, particularly the smaller and thinner landslides that likely experience larger basal pore‐water pressure changes. Based on climate model predictions for the next century in California, which include increases in average annual precipitation and increases in the frequency of dry‐to‐wet extremes, we hypothesize that there will be an overall increase in landslide activity.