Shear waves play a key role in seismic discrimination between explosions and earthquakes due to their different source mechanisms. However, shear waves are often observed in field explosions with unexpectedly large amplitude, and their generation mechanism is still a significant unresolved question in seismology. Many explanations have been proposed, including the asymmetry of explosive sources, and heterogeneity and/or anisotropy of the Earth’s subsurface. However, it has not been well understood whether source or velocity structure can independently and sufficiently explain the shear waves generated by explosions. Theoretically, tangential SH waves can be converted and scattered from vertical and radial SV waves due to anisotropy and heterogeneity. Thus, it is essential to understand the generation of SV waves by explosions. In this study, we utilize the frequency–wavenumber algorithm and 1D layered velocity models to simulate waveforms of isotropic explosions and double‐couple earthquakes at local distances (<20 km). Our results suggest that explosions and earthquakes may generate comparable SV waves if both occurred within a near‐surface velocity gradient zone. The earliest SV waves by explosions appear to originate from the near‐source region. It implies that P/SV amplitude ratios of explosions and earthquakes could be indistinguishable under certain circumstances.

Seismic discrimination between explosions and earthquakes is one of the most important missions in seismology, which was highlighted by the discrimination of underground nuclear explosions and tectonic earthquakes (Richards and Zavales, 1990; Bowers and Selby, 2009). Based on the differences in source location, geometry, and volume, three broad categories of seismic methods, that is, source depth, proportion of compressional and shear waves, and frequency content of specific phases, have been adopted for discriminating between explosions and earthquakes (Richards and Zavales, 1990). Direct source depth identification, based on either seismic data or remote sensing techniques, is relatively reliable but challenging (Bowers and Selby, 2009). Instead, waveform characteristic‐based discriminations have been widely adopted in the past decades. For example, the difference between body‐ and surface‐wave magnitudes (i.e., Ms:mb; e.g., Selby et al., 2012), the difference between local and coda duration magnitudes (i.e., MLMC; e.g., Koper et al., 2008, 2021), the amplitude ratio (e.g., Taylor et al., 1989; Walter et al., 1995) or spectral ratio (e.g., Fisk, 2006) of compressional and shear waves (e.g., Pg/Sg, Pn/Sn, Pg/Lg, Pn/Lg). However, the discrimination criteria of the aforementioned methods may be subjective in different studies, though the discrimination performance can be improved using network‐averaged results (e.g., Pyle and Walter, 2019; Wang et al., 2020) and distance corrections (e.g., Kim et al., 1997), especially for the low‐yield event discrimination at local distances (e.g., <50 km in O’Rourke et al., 2016; <150 km in Koper, 2020;,Koper et al., 2021; <200 km in Pyle and Walter, 2022). The classification discrepancy is highlighted by the debate on the seismological evidence for the existence of a low‐yield nuclear test by North Korea in May 2010 (Schaff et al., 2012; Ford and Walter, 2015; Zhang and Wen, 2015; Kim et al., 2017).

A big challenge in seismic discrimination is to understand shear wave generation by explosions, which is also one of the main goals in the recent Source Physics Experiments (SPE; Snelson et al., 2013). An isotropic explosion is not expected to generate significant shear waves, though their existence has long been observed and recognized in field explosions (e.g., Massé, 1981; Myers et al., 1999; Rodgers et al., 2010; Baker et al., 2012; Pitarka et al., 2015; Mellors et al., 2021). Multiple mechanisms have been proposed to explain the generation of shear waves by explosions, for example, asymmetry of the explosive source, spall closure, Taylor instability, nongeometrical phenomena, tectonic release, wave conversion on free surface or interfaces with velocity contrast such as S* (a strong nongeometric arrival; see definition in Hron and Mikhailenko, 1981) and pS, heterogeneity, anisotropy, and preexisting faults and joints (e.g., Johnson and Sammis, 2001; Baker et al., 2012; Xu et al., 2014; Pitarka et al., 2015; Hardage and Wagner, 2018; Vorobiev, 2023; and their cited papers). More recently, a number of studies have advanced our understanding of shear wave generation through waveform simulation of explosions in the SPE experiments (e.g., Pitarka et al., 2015; Steedman et al., 2016; Vorobiev et al., 2018; Mellors et al., 2021; Scalise et al., 2021; Vorobiev, 2023). For instance, Pitarka et al. (2015) suggested that wave‐scattering effects alone cannot fully explain shear wave generation, and that shear waves are generated at or very near the source. Mellors et al. (2021) found that waveforms are not consistent with a pure isotropic explosion, and the observed S waves originate from very near the source region through modeling of an explosion source with partial double‐couple added. However, it has not been fully investigated if either source or velocity structure independently and sufficiently explains the shear wave generation by explosions.

Tangential shear waves (e.g., SH waves) can arise due to many mechanisms, including prestress or tectonic release and nonspherical symmetry of shallow explosions due to the presence of the free surface, or the presence of multiple explosions in ripple‐fired mining shots. These waves can be converted and scattered from vertical and radial shear waves (e.g., SV waves) due to anisotropy and heterogeneity (Kennett and Mykkeltveit, 1984; Maupin, 1990; Baker et al., 2012). Thus, it is essential to first understand the generation of SV waves by explosions. In this study, we focus on a local blast and a nearby shallow earthquake, estimate P/S amplitude ratios from simulated waveforms of the two events, and demonstrate that the existence of a near‐surface velocity gradient zone may be sufficient to make earthquakes and explosions generate indistinguishable P/SV amplitude ratios on vertical components.

A pair of small intraplate earthquakes hit Dartmouth, Nova Scotia, Canada, in early March 2020, whose broadband waveforms are only recorded by a local seismic station. Zhang et al. (2021) located the two events through single‐station waveform modeling based on a regional velocity model (i.e., near surface + half‐space), which was constructed via waveform fitting of a nearby ground‐truth quarry blast (Mw 1.9) between its observations and synthetics of low‐frequency surface waves and high‐frequency body waves (Fig. 1). The two earthquakes are located at a recently constructed commercial development, and the depths are determined at 0.7 km with an uncertainty of <0.1 km (i.e., search interval) based on waveform modeling and Rg/S amplitude ratios (see details in Zhang et al., 2021). Such shallow source depth explains the “really loud bang” reported by residents (see Data and Resources). Detailed velocity model building and source parameter determination can be found in Zhang et al. (2021).

In the previous study, we observed that the blast generates significant shear waves including SV waves and SH waves in a frequency range of 7–20 Hz and surface waves in a frequency range of 2–6 Hz (Zhang et al., 2021; Fig. 1). With an isotopic explosive source and a well‐constructed 1D layered velocity model, SV and Rg waves can be well explained by the waveform simulation (Zhang et al., 2021). In this study, we conducted comprehensive waveform simulations to understand the generation of SV waves by explosions and investigate the effectiveness of the P/SV amplitude ratio method on seismic discrimination. We focus on the blast and one of the repeating earthquakes with larger magnitude for waveform simulation and seismic discrimination (Table 1; Zhang et al., 2021).

We simulate vertical and radial waveforms of the blast and the earthquake using a 1D layered velocity model and the frequency–wavenumber (fk) method and a triangular source time function with a duration of 0.05 s (Zhu and Rivera, 2002). The event‐station distance is 11.51 km for the blast and 7.15 km for the earthquake (Fig. 1). The depth of the open‐pit mining blast is typically assumed to be 0.01 km, and the depth of the earthquake was determined at 0.7 km (see details in Zhang et al., 2021). An isotropic explosive source is adopted for the blast, and an assumed vertical dip‐slip double‐couple source is used for the earthquake, which fits observations better than the other two fundamental focal mechanisms: strike slip and 45° dip‐slip (Zhang et al., 2021). The 1D layered velocity model is from Zhang et al. (2021), consisting of a near‐surface zone with velocity gradients and a half‐space layer. The VP and VS of the half‐space layer are 6.15 km/s and 3.55 km/s, respectively. The thickness of the near‐surface zone is determined at 0.21 km with 42 layers of 5 m thickness via waveform fitting of body and surface waves between their synthetics and observations (see details in Zhang et al., 2021). From top to bottom, VP and VS linearly change from 3.46 km/s to 6.15 km/s and from 2.0 km/s to 3.55 km/s, respectively (Fig. S1, available in in the supplemental material to this article). Typical high attenuations of Qp=100 and Qs=50 are adopted for the near‐surface zone, and Qp=1350 and Qs=600 for the half‐space layer (Zhang et al., 2021; Fig. S1).

We simulate waveforms of explosions and earthquakes at a range of source depths (i.e., 0.01 km and from 0.05 to 1 km with an interval of 0.05 km; Fig. 2), which cover the source depth range of the two events (i.e., 0.01 and 0.7 km). We follow Zhang et al. (2021) to filter waveforms in frequencies of 2–6 Hz and 7–20 Hz to separate the main surface waves and body waves, respectively. All the waveforms are normalized and aligned by their P arrivals. Here, we evaluate the waveform matching by comparing the waveform envelope and the ratio of seismic phases at various frequency bands rather than absolute amplitude differences due to potential biases from velocity heterogeneity and assumed sources (i.e., a perfect explosive source for the blast and a perfect double‐couple source with an assumed focal mechanism for the earthquake) in the waveform modeling. Waveform fits for the blast are better than the earthquake likely due to its relatively simple source mechanism. Overall, synthetic surface waves (2–6 Hz) and body waves (7–20 Hz) of the blast and earthquake approximately match well with the observations (Fig. 2), which verifies that our constructed 1D layered velocity model is an adequate representation of the velocity structure. At different source depths, simulated waveforms for the blast and earthquake show different characteristics (Fig. 2). In the next section, we will focus on their simulated waveforms with the same source depth for seismic discrimination, for which various focal mechanisms are considered for earthquakes.

We systemically investigate the seismic discrimination performance of the P/SV amplitude ratio method. We adopt the epicentral distance of the blast to simulate waveforms of explosions and earthquakes with different source depths (i.e., 0.01 km and from 0.05 to 1 km with an interval of 0.05 km; Fig. 3). A pure isotropic explosion source is adopted for explosions. To consider various focal mechanisms for earthquakes, we grid search their strike, dip, and rake with an interval of 20°, leading to 1805 double‐couple focal mechanisms. We compute the P/SV amplitude ratio for body waves of explosions and earthquakes on the vertical component, which is widely adopted in field seismic discrimination (e.g., Kim et al., 1993; Zhao et al., 2008), especially when only vertical geophones are available (e.g., Wang et al., 2020). Waveforms are filtered in a frequency range of 7–20 Hz to eliminate surface waves. Within time windows of P and SV phases, the root mean square of the maximum and the minimum amplitudes is adopted for amplitude estimation of synthetics and observations. For a fair comparison, we compare the P/SV amplitude ratios for simulated explosions and earthquakes with the same source depths. The P/SV amplitude ratios for both earthquakes and explosions fluctuate with change of source depths in the near‐surface zone but become relatively stable if their source depths are larger than the thickness of the near‐surface zone.

The results show that explosions and earthquakes tend to be confidently separated based on their P/SV amplitude ratios if they occur below the near‐surface zone (Fig. 3). However, the P/SV discriminations can be unstable or even fail if they occur within and close to the near‐surface zone; their P/SV amplitude ratios are very close or even highly overlapped (especially for source depths near the surface, e.g., ≤0.1 km in Fig. 3). Similar analysis on different frequencies (7–12, 10–15, and 13–20 Hz) do not result in significant changes (see Figs. S2–S4). We also evaluated the P/SV discrimination using another two near‐surface velocity gradient zones with thicknesses of 0.51 and 0.81 km, for which we use the same starting and ending velocities for the near‐surface zone as before (Fig. S1), leading to different velocity gradients in the near‐surface zone. The P/SV amplitude ratios of simulated explosions and earthquakes show similar distributions, resulting in consistent conclusions (Figs. S5, S6). These results suggest that it might not be stable or even impossible to discriminate an explosion from an earthquake using the P/SV amplitude ratio method if both occur within a near‐surface velocity gradient zone (especially near the surface).

The generation of shear waves from explosions is a significant unresolved question in seismology (Pitarka et al., 2015). It can be explained by imperfect explosive source and/or complex velocity structures (O’Rourke et al., 2016; Hardage and Wagner, 2018). This study demonstrates that, given the assumption of an isotropic explosive source, SV waves can be generated and propagated in a near‐surface velocity gradient zone (Fig. 4). Heterogeneity, anisotropy, tectonic release, and preexisting faults and joints are not included in our 1D layered velocity model. Thus, the generation of shear waves can be explained by only invoking near‐surface interfaces with velocity contrasts and a free surface, although some combination of the other complications mentioned earlier might also be involved in the generation of shear waves by explosions.

We observe that, like earthquakes, the S first arrivals of explosions are almost identical to their theoretical arrivals. Therefore, such shear waves unlikely come from the S*, which is characterized by lower velocity (Hardage and Wagner, 2018). Similarly, the free‐surface conversion pS may not be the reason as well, because (1) larger source depths result in earlier S first arrivals (see Fig. S7 and its caption) and (2) deeper stations below the velocity gradient zone record earlier S first arrivals (see Fig. S8 and its caption). In addition, we observe that strong downgoing shear waves propagate below the velocity gradient zone (see Fig. S8; similar observations can be found in Pitarka et al., 2015). Thus, the earliest shear waves originate from the near‐source region that is consistent with the observations of Pitarka et al. (2015) and Mellors et al. (2021), which may be caused by wave conversion near the source (Salvado and Minster, 1980), or nongeometrical phenomena (Babich and Kiselev, 1989), or Taylor instability (Wright and Carpenter, 1962) due to velocity gradients. The investigation of them is above and beyond the current scope of this study, which requires more waveform simulations and lab experiments.

Our further simulation suggests that a constant low‐velocity layer cannot generate similar shear waves as the field observations (e.g., amplitude and frequency content of S waves for the blast; see Fig. S9). Thus, the existence of a complex near‐surface zone with velocity gradients is sufficient for waveform simulation in this specific study, which may be characterized by either 3D velocity heterogeneity (e.g., Scalise et al., 2021) or 1D multiple gradient layers (this study).

The vertical‐component P/S amplitude ratio has been successfully and widely adopted for seismic discrimination between explosions and earthquakes in the past decades (e.g., Kim et al., 1993; Zhao et al., 2008; Wang et al., 2020), though average values on three components result in more stable results (e.g., Kim et al., 1997). Our results suggest that such success of P/S discrimination is likely attributed to source depth difference (i.e., earthquakes have larger source depths than explosions) rather than the source mechanism itself (i.e., double‐couple source vs. explosive source), similar to the Lg/P discrimination (Baker et al., 2004). With identically shallow source depths, we simulate waveforms of explosions and earthquakes, and find that a 1D velocity model with a near‐surface velocity gradient zone can be sufficient to generate significant SV waves for explosions, leading to the failure of the P/SV amplitude ratio method in discriminating explosions from earthquakes (Fig. 4). A near‐surface velocity gradient zone widely exists as seismic velocities increase with depths, especially in sedimentary basins, whose thickness can be up to a few kilometers (Evenick, 2021), enabling to accommodate tectonic and/or induced earthquakes (e.g., in the Western Canada sedimentary basin; Atkinson et al., 2016). As a result, principally speaking, P/SV amplitude ratio may not be stable or applicable for seismic discrimination between explosions and earthquakes when both occur with identically shallow depths within a near‐surface velocity gradient zone, though such circumstances might not be often in field observations (a near‐surface zone typically exhibits relatively low velocities and is less common to accommodate earthquakes).

We adopted a triangular source time function with a duration of 0.05 s for the blast waveform simulation, which may differ from that in the field blasting, though the waveforms fit reasonably well (Fig. 2). The blasting operator reported that it was a ripple‐fired blasting (i.e., multiple shot holes were fired with short delays; the detailed design is not available). Perhaps it is not surprising that the field waveforms can be modeled by a single‐shot explosive source. Smith (1989) suggested that a high‐frequency band of >35 Hz (much higher than the highest frequency of 20 Hz in our body wave comparison) is required to discriminate ripple‐fired blasting at regional distances. The effect of short delay times (milliseconds) may be strongly attenuated during wave propagation and thus unobservable (Smith, 1989). To test the effect of source complexity, we simulate waveforms using a more complex source time function with a longer duration (i.e., a trapezoidal function with a duration of 0.1 s) and find that the waveform fits change little (see Fig. S10 and its caption). Thus, different source time functions likely have little effect on the P/SV analysis.

We simulate waveforms for explosions and earthquakes at a single station using 1D layered velocity models and investigate the effectiveness of P/SV amplitude ratio on the vertical component for seismic discrimination. We acknowledge that our experiment has a few limitations because of the availability of seismic stations and simplicity of velocity models. First, only single‐station analysis was conducted, because we do not have other available stations in the field observations to verify our waveform simulation. Instead of using multiple stations, we grid search focal mechanisms for earthquakes to simulate waveforms with various wave radiation patterns. Second, different epicenter‐station distances were not well tested. We simulate waveforms for the blast and earthquake with distances up to 20 km, and find that their waveform characteristics are similar (Fig. S11), whereas comprehensive analysis and comparison are recommended if there are available events and stations in the field observations. Third, we analyzed the P/SV discrimination for simulated explosions and earthquakes under the same source depths, but not for field data, which is limited by the observations. Fourth, we did not evaluate the P/SV discrimination for surface explosions, because only buried sources can be adopted in our waveform simulation. Fifth, tangential SH waves were not simulated due to the simple 1D layered velocity models, which can be investigated using complex 3D velocity models (e.g., add scatters and topography to the current 1D layered model). Accordingly, the three‐component recordings can be used for evaluating the performance of P/S discrimination.

In this study, we investigate the generation mechanism of SV waves for explosions using simulated local waveforms of a pair of blasts and earthquake, and discuss the effectiveness of the widely used P/SV amplitude ratio method for seismic discrimination between explosions and earthquakes. The key conclusions are as follows:

  1. The existence of a near‐surface velocity gradient zone could be sufficient to generate significant SV waves for explosions with an isotropic explosive source.

  2. The earliest shear waves by explosions appear to originate from the near‐source region rather than the wave conversion from the surface.

  3. The single‐component‐based P/SV amplitude ratio method may not be stable or applicable for seismic discrimination between explosions and earthquakes if both occur within a near‐surface velocity gradient zone, although such circumstances may be very uncommon in field observations.

Waveform data were downloaded from Earthquakes Canada (https://earthquakescanada.nrcan.gc.ca/, doi: 10.7914/SN/CN). The frequency–wavenumber (fk) synthetic seismogram package was used for waveform modeling (https://www.eas.slu.edu/People/LZhu/downloads/fk3.3.tar; Zhu and Rivera, 2002). News reports about the two earthquakes can be found at https://www.halifaxtoday.ca/local-news/magnitude-26-earthquake-jolts-parts-of-dartmouth-2131000 and https://www.halifaxtoday.ca/local-news/aftershock-shakes-parts-of-dartmouth-saysearthquakes-canada-2133739, respectively. The maps in this article were made by the Generic Mapping Tools (GMT; https://www.generic-mapping-tools.org/; Wessel et al., 2013). The Supplemental Material contains 11 supplemental figures to support the discussions in the main text. All websites were last accessed in March 2023.

The author declares that there are no conflicts of interest recorded.

This work is supported by the Natural Sciences and Engineering Research Council of Canada Discovery Grant (Grant Number RGPIN‐2019‐04297) and the Canada Foundation for Innovation. The author thanks Conrad Brothers Ltd. for providing information about the quarry blast. The author thanks Ruijia Wang for the prereview and comments. The author thanks the Editor‐in‐Chief Keith Koper, Associate Editor Steven J. Gibbons, Alan Kafka, and another two anonymous reviewers for their comments and suggestions that greatly improved the article.

This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Supplementary data