← Back to Gallery

A Feature-based Liquefaction Image Dataset for Assessing Liquefaction Extent and Impact

In the aftermath of an earthquake, data collection is an important part of the response and is used for both loss assessment and data curation for model development. For liquefaction impacts, post-earthquake data collection often relies on field investigations, which are usually spatially limited and incomplete. Field investigations may capture liquefaction spatial extent and resulting deformations, but not consistently. With the increased availability of high-resolution optical imagery and radar technologies like Synthetic Aperture Radar (SAR), researchers in recent years have employed a combination of field surveys and visual and automated mapping of liquefaction surface effects using remotely sensed data sources. The inconsistencies in how the field investigations and remotely sensed observations are mapped can impede liquefaction inventory development, in part due to the range in size of liquefaction features (cm’s to 10s of m’s) and the inconsistency of feature attributes. Prior liquefaction inventories are point-based and summarize liquefaction occurrence across a site (10s of m) as a single point and as a binary variable: liquefaction occurrence vs. non-occurrence. Visual assessment of surface effects and detailed spatial mapping demonstrate that not all liquefaction features are equivalent and liquefaction inventories would benefit from a measure of size as well as deformation. In this study, we provide a liquefaction image library that includes spatial polygons and labels for over 2000 liquefaction features with information such as the infrastructure impacted and the size of the features across multiple events: including significant liquefaction events such as Christchurch (2011) and Tohoku (2011), as well as minor liquefaction events such as Haiti (2010) and Puerto Rico (2020). Our study proposes a methodology and establishes an image-based liquefaction inventory using remotely sensed data like optical and SAR imagery that not only provides occurrence and non-occurrence labels, but also provides information on impact in terms of feature size, impacted infrastructure and deformation.

Session: Modeling, Collecting and Communicating Post-earthquake Hazard and Impact Information [Poster]

Type: Poster

Room: Evergreen Ballroom

Date: 4/20/2022

Presentation Time: 08:00 AM Pacific

Presenting Author: Christina Sanon

Student Presenter: Yes

Additional Authors