A near-real-time model for estimating probability of road obstruction due to earthquake-triggered landslides

Abstract

Coseismic landslides are a major source of transportation disruption in mountainous areas, but few approaches exist for rapidly estimating impacts to road networks. We develop a model that links the U.S. Geological Survey (USGS) near-real-time earthquake-triggered landslide hazard model with Open Street Map (OSM) road network data to rapidly estimate segment-level obstruction risk following major earthquake activity worldwide. To train and validate the model, we process OSM data for 15 historical earthquakes and calculate the average segment-level landslide hazard from the USGS model for each event. We then fit a multivariate adaptive regression spline model for the probability of road obstruction as a function of road segment length and landslide hazard, using a training and validation dataset derived from the intersections of road networks with earthquake-triggered landslide inventories. The resulting probabilistic model is well calibrated across a range of earthquake events, with estimated obstruction probabilities matching the relative frequency of potential road obstructions. The model runs quickly and is capable of producing road segment-level obstruction estimates within minutes to hours of a major earthquake. However, in near-real-time application, the accuracy of the obstruction estimates will be dependent on the quality of the ShakeMap shaking estimates, which often improves with time as more information becomes available after the earthquake. By providing a rapid first-order translation of landslide hazard into potential infrastructure impacts, this model helps provide emergency responders with tangible information on initial areas of concern.

The US Geological Survey ground failure product: Near-real-time estimates of earthquake-triggered landslides and liquefaction

“…As more infrastructure and GF loss data become available, more detailed, quantitative loss estimation may become possible. To start to address this issue, Wilson et al (2021) developed a method for using the GF product model outputs and Open Street Map road network data to estimate the probability of road obstruction due to landslides, though this has not yet been implemented in the near-real-time system. Similar approaches could be added for other infrastructure types, though this must be done with care, ideally in an aggregate sense and using high-quality validation data because the GF models are not appropriate for site-specific application due to their simplicity and coarseness.…”

Detecting Turning Relationships and Time Restrictions of OSM Road Intersections from Crowdsourced Trajectories

“…This project has been growing rapidly, with more than 8.3 million registered users, thus realizing a global, widely, and freely used road network dataset [2]. Specifically, OSM road networks can be used in spatial analysis such as constructing building databases and the evaluation of economic development [3,4] and road obstruction [5]. In addition, these road networks support spatial applications such as identifying traffic congestion [6,7] and intelligent traffic management [8].…”