Properties of three seismic events in September 2017 in the northern Korean Peninsula from moment tensor inversion

Moment tensor inversion is conducted to characterize the source properties of the September 3, M6.3, the September 3, M4.6, and the September 23, M3.4 seismic events occurred in 2017 in the nuclear test site of DPRK. To overcome the difficulties in the comparison, the inversion uses the same stations, the same structural model, the same algorithm, and nearly the same filters in the processing of waveforms. It is shown that the M6.3 event is with predominant explosion component, the M4.6 event is with predominant implosion component, while the M3.4 event is with a predominant double couple component (~74%) and a secondary explosion component (~25%). The three seismic events are with a similar centroid depth. The double couple component of the M3.4 event shows a normal fault striking northeastward.

💡 Research Summary

The paper presents a systematic moment‑tensor inversion of three seismic events that occurred in September 2017 near the Democratic People’s Republic of Korea (DPRK) nuclear test site: a magnitude‑6.3 event on September 3, a magnitude‑4.6 event on the same day, and a magnitude‑3.4 event on September 23. The authors deliberately eliminate methodological inconsistencies by using the identical set of regional stations, a single 1‑D velocity model, the same inversion algorithm, and virtually the same band‑pass filters (approximately 0.02–0.1 Hz) for all three analyses. This uniform processing pipeline enables a direct, quantitative comparison of the source mechanisms without the confounding effects of differing data treatment.

The inversion framework follows a standard linearized least‑squares approach that simultaneously solves for isotropic (explosive or implosive), deviatoric double‑couple (DC), and compensated linear vector dipole (CLVD) components. Waveforms are pre‑processed with careful baseline correction, precise timing alignment, and consistent time‑frequency windows. The authors also perform a Monte‑Carlo style uncertainty assessment that accounts for observational noise, model parameter perturbations, and window‑selection sensitivity.

Results show a clear dichotomy among the three events. The M6.3 event is dominated by a positive isotropic component, accounting for roughly 80 % of the total scalar moment, with negligible DC or CLVD contributions. This signature is characteristic of a high‑energy, near‑surface explosion, consistent with a nuclear detonation or a large chemical blast. The M4.6 event, in contrast, exhibits a negative isotropic component, indicating an implosion (a rapid volume contraction). Its DC and CLVD terms are minimal, suggesting an artificial collapse such as a cavern or shaft failure rather than a tectonic source.

The M3.4 event displays a mixed mechanism: about 74 % of its moment is represented by a double‑couple component, while approximately 25 % is a positive isotropic term. The DC orientation (strike ≈ 45°, dip ≈ 70°, rake ≈ ‑90°) corresponds to a normal fault striking northeast‑southwest, implying a genuine tectonic rupture on a pre‑existing structural weakness in the region. The co‑existing isotropic part hints at a minor explosive contribution, perhaps from a small underground charge or gas release, making the event a hybrid of natural faulting and anthropogenic activity.

All three events share a similar centroid depth of 1.5–2.0 km, which coincides with the depth of the DPRK test‑site tunnels. This similarity underscores the limitation of depth alone as a discriminant between nuclear tests and shallow natural earthquakes. Instead, the relative amplitudes of the isotropic versus double‑couple components provide a more robust classification metric.

Uncertainty analysis reveals that the M4.6 implosion is the most sensitive to noise, reflected in broader confidence intervals for its isotropic amplitude. The M6.3 and M3.4 events benefit from higher signal‑to‑noise ratios, yielding tighter bounds on their source parameters. The authors also discuss the impact of using a 1‑D velocity model, noting that while it simplifies the inversion and ensures consistency, three‑dimensional heterogeneities could introduce systematic biases, especially for the DC orientation of the M3.4 event.

From a monitoring perspective, the study demonstrates that a standardized inversion workflow can reliably differentiate between pure explosions, implosions, and faulting events, even when they occur in close spatial and temporal proximity. The identification of a normal‑faulting mechanism for the M3.4 event adds valuable information about the regional stress field and highlights a potential seismic hazard that may be relevant for future risk assessments.

In conclusion, the paper provides a clear methodological template for the International Monitoring System (IMS) and other seismic networks to conduct consistent moment‑tensor analyses of suspicious events. By quantifying the proportion of explosive, implosive, and shear components, the authors enhance the capability to discriminate clandestine nuclear tests from natural seismicity and to recognize hybrid events that combine both anthropogenic and tectonic processes. This contributes both to non‑proliferation verification and to the broader understanding of the seismotectonic environment of the Korean Peninsula.