Supervised Machine Learning of High Rate GNSS Velocities for Earthquake Strong Motion Signals

Abstract

High rate Global Navigation Satellite System (GNSS) processed time series capture a broad spectrum of earthquake strong motion signals, but experience regular sporadic noise that can be difficult to distinguish from true seismic signals. The range of possible seismic signal frequencies amidst a high, location‐varying noise floor makes filtering difficult to generalize. Existing methods for automatic detection rely on external inputs to mitigate false alerts, which limit their usefulness. For these reasons, geodetic seismic signal detection makes for a compelling candidate for data‐driven machine learning classification. In this study we generated high rate GNSS time differenced carrier phase (TDCP) velocity time series concurrent in space and time with expected signals from 77 earthquakes occurring over nearly 20 years. TDCP velocity processing has increased sensitivity relative to traditional geodetic displacement processing without requiring sophisticated corrections. We trained, validated and tested a random forest classifier to differentiate seismic events from noise. We find our supervised random forest classifier outperforms the existing detection methods in stand‐alone mode by combining frequency and time domain features into decision criteria. The classifier achieves a 90% true positive rate of seismic event detection within the data set of events ranging from MW4.8–8.2, with typical detection latencies seconds behind S‐wave arrivals. We conclude the performance of this model provides sufficient confidence to enable these valuable ground motion measurements to run in stand‐alone mode for development of edge processing, geodetic infrastructure monitoring and inclusion in operational ground motion observations and models.

Validation of Peak Ground Velocities Recorded on Very-high rate GNSS Against NGA-West2 Ground Motion Models

“…Moreover, the ground velocity observations can be used to appropriately window displacement time series or to select only those stations that should have a significant displacement signal and exclude those that are effectively noise, thus improving the precision and accuracy of the geodetic source models within G-FAST. For example, Dittmann et al (2022b) trained a random forest classifier to select only parts of the GNSS velocity time series that had earthquake related ground shaking and showed a true positive rate of roughly 90% for earthquakes greater than M5 and out to hypocentral distances greater than 1000 km for larger events. Finally, GNSS derived velocities can be used directly in early warning systems either in existing seismic algorithms that rely on velocity observations or through magnitude scaling.…”

Characterizing High Rate GNSS Velocity Noise for Synthesizing a GNSS Strong Motion Learning Catalog

“…Observations are weighted as a function of satellite elevation angle with a seven degree elevation mask. While development accommodating precise orbits (Shu et al, 2020), multi-GNSS, cycle slip detection/mitigation (Fratarcangeli et al, 2018), and higher order noise source mitigation is ongoing and warranted, the current method is capable of capturing ground motions of nearfield M4.9 and larger sources at teleseismic distances (Crowell, 2021;Dittmann et al, 2022b).…”

“…Machine learning (ML) models combine a range of feature inputs to improve the decision confidence in separating seismic signal from noise (e.g. Meier et al, 2019;Dittmann et al, 2022b) in stand-alone mode. However, the generalization performance of any such classifier or deeper ML model will ultimately be limited by the model selection and optimization, the extent of the labeled catalog for training, and the quality of the labels.…”

Interpolation of GNSS Position Time Series Using GBDT, XGBoost, and RF Machine Learning Algorithms and Models Error Analysis

“…In the last thirty years, more than 20,000 global navigation satellite system (GNSS) continuously operating reference stations (CORS) have been established worldwide, and the GNSS position time series observed by these CORSs can provide effective data support for geoscience research. By analyzing the GNSS position time series, researchers have studied crustal movement [1][2][3], the maintenance of regional or global geodetic reference frames [4][5][6], engineering deformation monitoring [7][8][9], and other geodynamic phenomena [10][11][12].…”