Exploring Generalized Relationships Between Rockfalls and Seismograms
Description:
Rockfall event parameters are typically estimated using field surveys or satellite imagery. These methods have poor temporal resolution, and potentially poor spatial resolution depending on the data and have uncertainties in the measurements that may be challenging for hazard characterization. The seismic waves generated by these rockfalls carry information about their source. Therefore, seismic monitoring of rockfalls provides an excellent opportunity to monitor rockfalls. By understanding the relationships between the time-frequency representation of the seismograms and the rockfall dynamics, one can estimate event parameters for the small to moderate sized slope movements for which event parameters cannot be estimated using traditional methods. We explore the relationship between over 300 features extracted from seismograms and three rockfall event parameters - volume, runout distance and drop height - for the rockfall events in IRIS Exotic Seismic Event Catalog (ESEC). IRIS ESEC contains a database of 52 rockfalls (as of 10 January 2023) compiled from published studies. These rockfalls have occurred in a wide range of geographical locations, and range in size from less than 100 m3 to more than 1,00,000 cubic meters.
Our primary motivation is to find features that show high correlation with rockfall event parameters. We use correlations and a random forest algorithm. Our preliminary results indicate that certain features characterizing the spectral content of the seismograms show high correlation with all the event parameters. Some of these features show even better correlation compared to some of the features previously chosen by other studies (Dammeier et al. 2014, Manconi et al. 2016, Hibert et al. (2017a, b)).
Session: Opportunities and Challenges for Machine Learning Applications in Seismology [Poster]
Type: Poster
Date: 4/19/2023
Presentation Time: 08:00 AM (local time)
Presenting Author: Akash Kharita
Student Presenter: Yes
Invited Presentation:
Authors
Akash Kharita Presenting Author Corresponding Author ak287@uw.edu University of Washington |
|
|
|
|
|
|
|
|
Exploring Generalized Relationships Between Rockfalls and Seismograms
Category
Opportunities and Challenges for Machine Learning Applications in Seismology