Slow slip events detection algorithm based on an optimization-based signal decomposition framework: A case study in Cascadia zone

Slow Slip Events (SSEs) are critical for understanding subduction zone tectonics and earthquake prediction; however their detection is challenged by low-magnitude-offsets and data gaps. To address these challenges, this paper introduces an optimization-based signal decomposition (OSD) framework capable of automatically processing signals with missing data. We applied and validated this framework with GNSS coordinate time series in the Cascadia subduction zone, benchmarking its performance against the existing SSEs catalog. The proposed high-magnitude-offset detection method achieved an accuracy of 67.21% in single-station SSE detection, significantly outperforming traditional methods such as the Relative Strength Index (RSI; 32.24%) and deep learning methods like bidirectional Long Short-Term Memory (bi-LSTM; 44.41%). Additionally, we proposed a complementary velocity-based screening strategy that successfully identified low-magnitude-offset SSEs and events obscured by data gaps. Through cluster analysis of single-station detection results, we successfully identified the spatiotemporal boundary of the majority of SSEs. Finally, we established an anomaly catalog for uncataloged period from 2018 to 2024, which further demonstrates the method's efficacy in characterizing the spatiotemporal features of SSEs. The OSD-based SSEs detection framework identified SSEs with diverse kinematic patterns using raw geodetic data, facilitating the construction of high-quality SSEs catalogs. These advancements enhance our understanding of subduction zone dynamics and provide a robust technical foundation for seismic hazard assessment.