Spatial Interactions Govern Tectonic Tremor Activities in the Nankai Trough: A Statistical Approach for Spatiotemporal Pattern Modeling and Anomaly Detection

Abstract Establishing a quantitative framework to evaluate the spatiotemporal patterns of slow earthquakes and to detect their anomalous activities is essential for understanding diverse slip behaviors on plate boundaries and seismic hazard assessment. In this study, we focus on deep tectonic tremors, which are one manifestation of slow earthquakes and are detectable through seismological observations along the Nankai Trough. We develop a probabilistic model based on a multivariate Hawkes process to describe both the temporal and spatial characteristics of their activity. Our analysis shows that more than half of the tremors are attributable to interactions with neighboring regions, which underlines the importance of incorporating spatial interactions into the forecasting model. Along‐dip and along‐strike variations in background tremor seismicity and the effective durations of inter‐tremor interactions revealed in our model are consistent with previous geophysical and geological observations and conceptual frameworks, including tremor migration, depth‐dependent slip modes, partial overlap with slow slip events (SSEs), and along‐strike variations in plate convergence rate. We further compare the model‐predicted tremor activity with observations and quantify their differences in cumulative event counts using the Kolmogorov–Smirnov test to identify transient anomalies. The detected anomalies include both short‐duration (∼0.1 days) and long‐duration (∼100 days) activations and quiescences. Although their spatial extent is much smaller (25 km) than that of SSEs, half of the short‐term anomalies partially correlate with geodetically detected short‐term SSEs. These results demonstrate the potential of our approach to provide a robust framework for forecasting slow earthquake activity and detecting its changes.