Near‐Real‐Time Modeling of Landslide Impacts to Inform Rapid Response: An Example from the 2016 Kaikōura, New Zealand, Earthquake

Abstract

Reliable methods for the near‐real‐time modeling of landslide hazard and associated impacts that follow an earthquake are required to provide crucial information to guide emergency responses. After the 2016 Kaikoura earthquake in New Zealand, we undertook such a near‐real‐time modeling campaign in an attempt to pinpoint the location of landslides and identify the locations where roads and rivers had been blocked. The model combined an empirical analysis of landslide hazard (based on previously published global relationships) with a simple runout model (based on landslide reach angles). It was applied manually, with a first iteration completed within 24 hrs of the earthquake and a second iteration (based on updated shaking outputs) within ∼72 hrs . Both models highlighted the expectation that landsliding would be widespread and that impacts to roads likely meant that the township of Kaikoura was cut off from the surroundings. These results were used by responders at the time to formulate aerial reconnaissance flight paths and to identify the risk that landslide dams could cause further damage. Subsequent model verification based on available landslide inventories shows that although these models were able to capture a large percentage of landslides and landslide impacts, the outputs were overpredicted, limiting their use for pinpointing the precise locations of triggered landslides. To make future versions of the model more useful for informing emergency responses, continued work must be done on modification and adaptation of the approach so that this overprediction is minimized. Nevertheless, the results from this study show that the model is a promising initial attempt at near‐real‐time landslide modeling and that efforts to automate the approach would greatly increase the utility and speed of modeling in future earthquakes.