← Back to Gallery

Session: Empirical Modeling of Site Effects [Poster]

Type: Poster

Date: 10/12/2023

Time: 07:00 AM

Room: Stanley Park Ballroom

Classification Algorithms for Sedimentary Basins in Southern California

Seismic site response is influenced by a variety of physical mechanisms in which amplification due to topography and three-dimensional features from sedimentary basins plays a crucial role. Contributions from sedimentary basins to site response is referred to as basin effects which includes the generation of propagating surface waves due to trapped (reflected and refracted) seismic waves, focusing of seismic energy due to basin shape, and resonance of the entire basin sediment structure. Current ground motion models (GMMs) consider basin effects in their estimates of ground motion hazard but this requires an indication of if the location of interest is within or outside a basin. This is relatively straightforward for a few sites but non-trivial for a regional study. As such, there is a need to systematically classify sites as basins or non-basin. In this study a quantitative classification algorithm is developed for southern California that estimates the probability of a site residing within a basin. This was accomplished by exploring the relationship between geomorphological features such as elevation, slope, curvature, and surface texture, with unambiguous classifications of basin and non-basin sites using a suite of machine learning methods. Feature engineering analysis involving Principal Component Analysis revealed surface texture as the optimal identifier for the classification model development. Performance analysis unveiled three models as the best classifiers of basin and non-basin sites. These three models are derived from logistic regression, support vector machine, and extreme gradient boosting, and as such their combined estimates capture the epistemic uncertainty associated with varying modeling approaches. The resulting basin/non-basin classification algorithm suite provides probability maps that will serve as a necessary tool for forward application of seismic hazard assessments that incorporates region specific basin effects (i.e., the U.S Geologic Survey National Seismic Hazard Model).

Presenting Author: Rashid Shams

Additional Authors