Rapid and Temporary Deployment of Fiber‐Optic Arrays to Measure Surface Chemical Explosion Series with Distributed Acoustic Sensing Open Access

Temporary seismic instruments are often rapidly installed to measure the ground motions of events of interest. These deployments typically occur in response to unpredictable natural events, such as major earthquakes (e.g., Meltzer et al., 2019; Cochran et al., 2020) and volcanic eruptions (e.g., Wu et al., 2020), or in anticipation of known events, such as the re‐entry of space capsules (Yamamoto et al., 2011). In these scenarios, deployment speed and spatial coverage are critical for effectively capturing the seismic wavefield and enhancing our understanding of the associated physical processes.

Although traditional geophones can be deployed quickly, acquiring thousands of units in a short time frame (i.e., a few days to weeks) can be challenging, often resulting in deployments with insufficient spatial coverage. In this context, distributed acoustic sensing (DAS) emerges as a cost effective and attractive alternative to dense geophone deployments. DAS technology uses fiber‐optic cables, employing repeated laser pulses sent down the fiber to achieve meter‐scale measurements of ground vibrations (i.e., strain or strain rate) that conventional seismometers cannot match (see Hartog, 2017 for a comprehensive review of DAS technology). However, our understanding of the wavefield captured along fiber‐optic cables is still evolving compared to that of geophones, and further investigation is needed.

Temporary DAS deployments have been successfully conducted in response to major earthquakes for analyzing aftershock sequences (Karrenbach et al., 2020; Li et al., 2021), studying volcanic activity (Jousset et al., 2022), and recording the re‐entry of a space capsule (Carr et al., 2024; Silber et al., 2024) and explosions (Abbott et al., 2019; Mellors et al., 2022; Viens and Delbridge, 2024). Although some deployments leverage existing telecommunication fibers for efficient data collection (e.g., Li et al., 2021), labor‐intensive deployments of fiber‐optic cables may be necessary in remote areas lacking telecommunication infrastructure (e.g., Jousset et al., 2022). Moreover, the geometry of existing telecommunications cables may not always be optimal for capturing the seismic wavefield generated by events of interest.

In this study, we present a case study of two rapid and temporary DAS deployments conducted near Socorro, New Mexico, United States, to capture ground motion from two series of surface chemical explosions on 8 May and 16 October 2024. Each series consisted of four 1‐ton trinitrotoluene (TNT)‐equivalent explosions, followed by a 10‐ton TNT‐equivalent explosion. The smaller explosions were conducted in pairs with a 3‐min interval between explosions, and the larger blast occurred several hours later, several kilometers west of the first explosions (Fig. 1a).

Both deployments, led by a collaborative team from the Los Alamos National Laboratory and the New Mexico Institute of Mining and Technology, involved separate fiber‐optic cable installations. For the October series, we also deployed collocated geophones and anemometers to facilitate direct comparisons between DAS and geophone data, as well as assess the impact of fiber burial. This setup offers a unique opportunity to analyze repeated surface explosions of varying sizes, offering additional validation of DAS for explosion monitoring. The article outlines the methods and materials used in the two DAS campaigns and offers a “lessons learned” overview to guide future rapid, temporary DAS‐based experiments. Finally, we present preliminary results, focusing primarily on cable coupling effects and comparisons with geophone data.

For both explosion series, we deployed a fiber‐optic cable at a site located in the dry riverbed known as the Arroyo de Los Pinos (ADLP). This location was chosen for its proximity to the explosion series—about 12 km to the east (Fig. 1a) —and its accessibility through the New Mexico Institute of Mining and Technology. This site hosts a sediment transport laboratory (STL, Fig. 1b), dedicated to studying sediment movement during flash floods that typically occur during the summer months of the United States southwest monsoon season (Stark et al., 2021; Luong et al., 2024). In the summer of 2024, five flash floods were documented, including one on the evening of 21 July, which reached water levels of approximately 1.35 m and had a peak discharge of approximately $56m3/s$⁠, altering the morphology of the riverbed between the two explosion series.

For the May and October field experiments, our team deployed approximately 2 km of fiber‐optic cable from a spool containing 4.5 km of tight‐buffered fiber, encased in a polyurethane jacket reinforced with aramid yarn. Both deployments were conducted on the day preceding each explosion series.

The dry riverbed at the experiment site facilitated vehicle access for fiber deployment. The spool was mounted on a dispenser secured to the back of a pickup truck, allowing the cable to be deployed as the truck moved along the riverbed. This operation required a team of three: one driver, one person in the back unspooling the fiber, and a third person walking behind to ensure the unspooling process ran smoothly and to position the fiber appropriately. Maintaining proper tension on the cable was critical; too much tension could stress the cable and lead to breakage, whereas too little could result in the cable getting caught in the spool dispenser. We laid approximately 1950 m of fiber on the ground in less than 2 hr before embarking on the challenging task of burying it beneath a thin layer of rocks and sediment (Figs. 1b, 2). Burying the fiber using picks and shovels proved difficult due to the compact nature of sediments and rocks in the dry riverbed. It took our team of four scientists five hours to trench and bury 1 km of fiber under a few centimeters of sediments.

The layout of the fiber deployment was constrained by geographical and physical boundaries. The elevated level of the Rio Grande in May (slightly higher than shown in Fig. 1b) and a fence bounded the west‐southwest and east‐northeast ends of the fiber deployment. The northernmost point of the deployment was also fenced off. From the DAS interrogator unit (IU), the fiber extends into the ADLP dry riverbed and runs west‐southwest toward the Rio Grande. Upon reaching the river, the fiber loops back up the ADLP riverbed, heading east‐northeast until it encounters the fence. It then traces a back‐and‐forth path to the Rio Grande before being laid above ground along the west‐southwest dirt road adjacent to the ADLP riverbed, eventually returning near the DAS IU. The fiber then runs north‐northwest along the north‐northwest–south‐southeast dirt road until reaching the northernmost point, before returning back to the DAS IU. The remaining ∼2.5 km of fiber is left on the spool to act as a microphone (Zamarreño et al., 2017; Ma et al., 2019; Chambers and Shragge, 2023; Silber et al., 2024) and capture acoustic and infrasound signals from the explosion series. The buried and unburied sections of fiber are shown in Figure 1b. Figure 2b,c shows the buried fiber at two locations in the ADLP dry riverbed, and Figure 2d shows the surface fiber from the northernmost point looking south‐southeast, highlighting the dense tree coverage flanking both sides of the north‐northwest–south‐southeast dirt road.

A three‐component broadband seismometer deployed by the New Mexico Institute of Mining and Technology also recorded continuous data during the May explosion series with a 1000 Hz sampling rate (Fig. 1c). A battery issue prevented the seismometer from recording during the October explosion series.

For the second fiber deployment, we used a hitch‐mounted trencher to dig a shallow trench while simultaneously unspooling the fiber from the truck. The trencher was attached to the truck with a 15 cm (6‐inch) adapter, allowing for a perforation depth of approximately 10 cm (Fig. 3b). After digging the trench, the cable was placed inside and covered. This approach significantly reduced physical labor, enabling us to bury the same length of fiber (i.e., ∼1 km) in just two hours with the same number of team members. In addition, we buried 5 three‐component and 2 one‐component (vertical) 5 Hz SmartSolo geophones at a shallow depth of about 5 cm. The geophones recorded the velocity wavefield with a sampling frequency of 1000 Hz, and their locations are marked in Figure 1c. The geophones were leveled, and the horizontal components of the three‐component sensors were oriented in the standard configuration (i.e., to true north).

In October, the water level of the Rio Grande level was lower (i.e., lower than on the satellite view of Fig. 1c), which allowed us to extend the fiber deployment farther south (Fig. 1c). The eastern and northern ends of the fiber deployment remained the same as in May. From the DAS IU, the fiber extends into the ADLP dry riverbed and runs southwest toward the Rio Grande and the N0 geophone. Upon reaching the river, the fiber turns south to follow the Rio Grande course toward the N2 geophone. At the southernmost point, it reverses direction and heads back north toward the ADLP dry riverbed, continuing east‐northeast passing the N3–N5 geophones until it reaches the fence. The fiber is then laid above ground along the west‐southwest dirt road heading back toward the DAS IU. A back and forth along the north‐northwest–south‐southeast dirt road is then performed before returning in the ADLP dry riverbed near the DAS IU.

We finally laid approximately 220 m of unburied fiber in the ADLP dry riverbed next to the buried fiber to quantitatively assess the effect of fiber burial. We deployed 110 m of unburied fiber starting near the STL and extending toward the west‐southwest until it reaches the turnaround point shown in Figure 3d. The fiber then returns toward the STL still unburied, but with rocks placed every ∼2.5 m on top of the fiber (Fig. 3d). The remaining ∼2.3 km of fiber was finally left on the spool on the northern bank of the dry riverbed.

We used a Silixa iDASv2 IU to record continuous DAS data. In both experiments, data were recorded over 2500 m of fiber with a spatial sampling of 1.021 m, a fixed gauge length of 10 m, and a temporal sampling rate of 1000 Hz. For both deployments, continuous data were recorded for several hours on the day before the explosion series during fiber deployment and burial. It was then restarted on the days of the May and October explosion series, recording continuously for 8 and 6 hr, respectively.

The relatively remote location of the deployment site precluded access to a power outlet for supplying power to the DAS IU. For both deployments, we utilized a BLUETTI Power Station AC300 connected to a B300K expansion battery, providing a total capacity of ∼2700 watt‐hour (Wh). The DAS IU and monitor consumed an average of 260 Wh while collecting data, ensuring approximately 11 hr of operation. In addition, two solar panels were connected to the power station, generating up to 500 Wh during full sun periods, which could extend the experiment’s potential duration to over ∼15 hr even in cloudy conditions. Finally, it is also worth noting that the expansion batteries can be connected in series. For instance, connecting the BLUETTI Power Station AC300 to four B300K batteries would yield a maximum capacity of ∼11,000 Wh, enabling DAS data recording for over 42 hr without reliance on solar panels.

The removal of the fiber‐optic cable for both deployments was conducted in a similar manner. After shutting down the DAS IU, we removed the buried fiber from the ground and began respooling it onto the reel that was reinstalled on the spool dispenser on back of the truck. The entire cleanup process took less than two hours, with our team of three completing the respooling of the 2 km of fiber in under one hour. Although not directly applicable to this experiment, we strongly advise researchers against creating loops of fiber to bring the cable closer to the spool, as this can lead to tangling, making it extremely difficult to untangle later.

The high sample rate (1000 Hz) and spatial sampling (1.021 m) for the two experiments produced a total of 373 GB of DAS data. For each of the 10 surface chemical explosions, 120 s of strain‐rate $ε˙(t)$ data, starting 20 s before each detonation, along the fiber with a 200 Hz sampling rate can be retrieved through the website available in Data and Resources along with the decimated (200 Hz) geophone data set for the October series. Each data file contains the required metadata information from the DAS metadata standard (Lai et al., 2024).

In the May and October experiments, we recorded the seismoacoustic waves generated by both the 1‐ and 10‐ton TNT‐equivalent chemical explosions. The strain‐rate waveforms from the first 1‐ and 10‐ton explosions for both deployments are presented in Figure 4. The strain‐rate data are band‐pass filtered between 0.5 and 100 Hz using a two‐pass, four‐pole Butterworth filter. In addition, a median waveform, calculated from all channels at each time point, is subtracted from each individual trace to minimize potential laser noise.

For the first 1‐ton TNT‐equivalent explosions, the first acoustic waves, which travel with a velocity of ∼330 to 390 m/s, arrive approximately 31 s after detonation and are clearly visible in Figure 4a,b. In contrast, the first seismic waves, which should arrive around 4 s after detonation based on a P‐wave velocity of 3000 m/s, are barely visible. This can be attributed to the surface nature of the chemical explosions, typically generating less seismic energy compared to buried explosions, combined with the distance of about 12 km from the explosion site. In contrast, the 10‐ton TNT‐equivalent explosions produced clearly visible seismic and acoustic waves (Fig. 4c,d), which first arrive approximately 4 and 36 s after detonation given their 3000 and 330–390 m/s velocities, respectively. The 5 s difference in acoustic wave arrival time between the 1‐ and 10‐ton explosions is due primarily to the varying distances, with the 10‐ton TNT‐equivalent explosion being located farther away from the DAS array (Fig. 1a).

In Figure 4, the buried and unburied sections of fiber can be clearly identified, with the unburied sections displaying higher levels of noise, generated primarily by wind. For the May experiment, the first 150 m of fiber and the segment between 1250 and 1450 m from the IU are unburied, located in the dry riverbed and on the west‐southwest dirt road, respectively. The section between 1450 and 1950 m runs along the north‐northwest–south‐southeast dirt road, benefiting from tree coverage that significantly mitigates wind noise (Fig. 2d). For the October series, the first 75 m and the section between 1175 and 1375 m from the IU are unburied and located on the side of the dry riverbed and on the west‐southwest dirt road, respectively. The unburied section between 1375 and 1875 m is located along the north‐northwest–south‐southeast dirt road and also benefits from the tree coverage to mitigate wind noise. The fiber sections between 1900–2010 m and 2010–2120 m from the IU are the unburied and unburied with rocks sections, respectively, and are collocated with the buried section that extends from 75 to 185 m from the DAS IU. A clear difference of noise levels can be observed between these three different types of fiber installation that is further quantified in the Buried, unburied with rocks, and unburied fiber data comparison section.

The strain‐rate data recorded along the Rio Grande, from 300 to 750 m from the DAS IU during the October explosion series, exhibit different behavior compared to the ADLP dry riverbed data (Fig. 4b,d). Notably, the first acoustic wave arrival is less distinct and appears more scattered than along the dry riverbed. In addition, we observe longer duration ground motions along the Rio Grande than in the ADLP dry riverbed. During fiber deployment, we noted that the sediment characteristics along the Rio Grande were markedly different from those in the ADLP dry riverbed, consisting of softer, dry clay sediments with almost no rocks in the upper 30 cm. The fiber was also more challenging to remove from the ground in this section. These subsurface differences, along with the different fiber orientation capturing another component of the wavefield, may explain the observed variation in the recordings, though further investigation is required to fully understand the underlying causes.

For the May and October first 1‐ and 10‐ton TNT‐equivalent explosions, we present the strain‐rate waveforms band‐pass filtered in the 0.5–4 Hz frequency range, recorded at the same location near the N5 geophone (Fig. 5a). The selected channels for these explosions are positioned at distances of 1224 m (May) and 1142 m (October) along the fiber, as highlighted in Figure 4c,d.

To quantify the similarity between the May and October first 1‐ and 10‐ton TNT‐equivalent explosions, we calculate the correlation coefficient (CC) between the strain‐rate waveforms over a 25 s seismic‐wave window. For the first 1‐ton TNT‐equivalent explosion, the CC is 0.56 between May and October, suggesting a good match between the seismic waves recorded with DAS, despite the relatively faint seismic signal (Fig. 4a,b) and potential minor positional differences due to fiber redeployment. For the 10‐ton explosion, the CC increases to 0.75, reflecting the stronger seismic signal from the larger explosion. However, notable differences emerge for the acoustic waveforms. Specifically, the initial acoustic arrivals in October are delayed by approximately 1 s compared to May. Furthermore, the amplitude of the acoustic waves from the first 1‐ton explosion in May is slightly higher than those from the 10‐ton explosions. These differences are likely caused by variations in atmospheric and wind conditions, as well as the greater distance of the 10‐ton TNT‐equivalent explosions from the recording site compared to the 1‐ton TNT‐equivalent explosions.

Figure 5b,d shows a comparison of the DAS acceleration waveforms with the N5 geophone acceleration data for the first 1‐ and 10‐ton TNT‐equivalent explosions in the 0.5–4 Hz frequency band. For the first 1‐ton TNT‐equivalent explosion, the CC values between the DAS data and the N5 geophone acceleration data over the seismic‐wave window are 0.41 and 0.68 for the May and October waveforms, respectively. The CC values for the 10‐ton TNT‐equivalent explosion are slightly higher, with values of 0.51 for May and 0.69 for October. In addition, we observe a good correspondence between the DAS‐converted acceleration amplitudes and the geophone acceleration levels for the seismic waves. However, for acoustic waves, the DAS‐converted acceleration amplitudes are higher than those measured by the N5 geophone. This discrepancy arises because the apparent velocity correction factor of 1000 m/s exceeds the 385 m/s apparent velocity of the acoustic waves recorded along the fiber. To match the acoustic wave amplitudes recorded by the N5 geophone, the amplitude of the DAS data should be corrected using an apparent velocity of 385 m/s.

In Figure 7a, we illustrate the effect of a 10 m gauge length, used with the Silixa iDASv2 for the experiment, across a range of frequencies and phase velocities. Clear drops of the DAS response can be observed at frequencies above 30 Hz for phase velocities lower than 1000 m/s. To further analyze the impact of gauge length effects on acoustic waves, we average the DAS response between phase velocities ranging between 330 and 390 m/s. The resulting averaged DAS response for acoustic waves is shown in Figure 7b together with the Fourier spectra of the acoustic waves recorded for the May and October 10‐ton TNT‐equivalent explosions. The Fourier spectra are computed from the waveforms recorded between 35 and 45 s after detonation, and are slightly smoothed with a moving average filter that runs over five frequency samples. The amplitude variations of the recorded data with frequency match remarkably well with the DAS response, highlighting the impact of gauge length effects on the recorded acoustic waves at frequencies higher than 30 Hz.

In Figure 8a, we compare the seismic waves in the 1–10 Hz frequency band from the October 10‐ton explosion for buried, unburied with rocks, and unburied DAS channels, all located approximately 92 m from the turnaround point, a few meters away from the STL location. We observe a good correlation in the phase of the seismic waves across the three types of deployed fibers, with the buried fiber exhibiting higher signal levels compared to the unburied fibers. The unburied with rocks channel also shows slightly higher seismic‐wave levels relative to the unburied channel.

We repeat the SNR measurements across 100 collocated channels, discounting the first and last five channels of the 110 collocated channels given the 10 m gauge length, and show the results in Figure 8b. The mean SNR for the buried fiber is 20 dB, which is more than twice that of the unburied with rocks (9 dB) and unburied channels (8.2 dB). Among the 100‐channel section, 61 out of 100 unburied with rocks channels exhibit higher SNR values than the unburied channels. This suggests that placing rocks on top of the fiber may provide a minimal improvement in SNR. However, we cannot entirely discount the possibility that the 50 cm difference in location between the two unburied fibers (Fig. 3d) contributes to these variations.

We also calculate the SNR for acoustic waves in the 1–50 Hz frequency band, with results presented in Figure 8c,d. A similar trend is observed for acoustic waves, with buried channels significantly yielding higher SNR levels. Across the 100 collocated channels, 65 out of 100 channels show higher SNR for the buried with rocks compared to the unburied channels. This comparison also underscores the benefits of burying fibers whenever feasible, even for recording acoustic signals. If burial is not possible, strategically placing rocks or sand bags every few centimeters along the fiber may help reduce noise levels, albeit with only minimal improvements.

Finally, we observe lateral variations of the SNR levels for both the seismic and acoustic waves, especially for the unburied and unburied with rocks sections. This can be explained by the fact that the fiber was not deployed perfectly straight along this 110 m section (i.e., Fig. 3d). The SNR variations follow the path of the fiber, which indicates that different types of waves may be recorded depending on the slight changes of fiber orientation.

To further understand the impact of wind noise on the buried and unburied DAS data, we also collected wind speed data every 5 s at the STL location (Fig. 1c), in the middle of the dry riverbed, using a BTMETER BT‐100APP anemometer. The instrument was positioned on the ground and aligned along the axis of the riverbed, facing the west-southwest direction. Although the anemometer does not capture wind direction, the prevailing wind at this location generally originates from the Rio Grande, and comes up the dry riverbed. To mitigate high-frequency fluctuations and enhance signal clarity, we applied a two-minute moving window smoothing to the wind speed data, which are presented in Figure 9 over a 5‐hr period on 16 October 2024.

To assess the impact of wind noise on the strain‐rate data recorded along the buried, unburied with rocks, and unburied fiber sections, we selected continuous data from 100 collocated DAS channel for each installation type. We then compute the power of the strain‐rate data within the 0.5–50 Hz frequency band for each channel. Power is defined as the average of the squared amplitudes over nonoverlapping 5 s windows, calculated over the 5‐hr data collection period. For each installation type, the strain‐rate power data from the 100 channels is averaged, and the resulting data are further smoothed using a two‐minute moving window to reduce the influence of transient signals, such as those caused by explosions or nearby human activity.

The strain‐rate power data from the unburied and unburied with rocks fiber sections, after channel averaging and two‐minute smoothing, show good correlations with the wind speed data, with CCs of 0.63 and 0.62, respectively (Fig. 9). In contrast, the CC between the buried fiber and the wind data is 0.06, indicating a lack of correlation. This further underscores the effectiveness of burying fibers, even under a thin layer of sediment, in reducing the impact of wind noise on measurements. In addition, the strain‐rate power amplitude for the unburied fiber consistently exceeds that of the unburied fiber with rocks, with average power values of $185×10−15s−2$ and $160×10−15s−2$⁠, respectively. The average power value for the buried fiber is significantly lower at $6×10−15s−2$⁠, further demonstrating the advantage of fiber burial in minimizing wind‐induced noise.

In this study, we detailed the temporary deployment setup of approximately 2 km of fiber‐optic cables to record DAS data from two series of surface chemical explosions in New Mexico, United States. Both deployments captured seismoacoustic waves from 1‐ and 10‐ton TNT‐equivalent explosions that occurred ∼12 km away. We showed that burying the fiber under a shallow layer of sediment significantly helps in reducing wind noise and capturing signals with high SNRs. Although fiber burial is by far the most labor‐intensive part of the DAS experiment, it can be made easier with the use of a hitch mounted trencher on a truck. We also showed that the recorded DAS data match well with collocated geophone data, although gauge length effects impact the high frequency (>30 Hz) content of the DAS data for acoustic waves.

Future research will focus on characterizing the chemical explosion sources using coda spectral ratio and beamforming techniques applied to both seismic and acoustic waves. In addition, infrasound modeling will be used to further investigate the characteristics and propagation of acoustic waves. In addition to the explosion series, the brief, yet continuous, nature of the collected data set provides an opportunity to investigate the subsurface shallow structure at the STL before and after the 2024 monsoon season. Ambient noise techniques may be employed to better understand the different nature of sediments between the Rio Grande riverbed and the ADLP dry riverbed, assess potential changes induced by flash floods, and infer the depth of the water table.

For each of the 10 surface chemical explosions, 120 s of DAS strain‐rate data along the fiber—downsampled to 200 Hz and starting 20 seconds before each detonation—are available at doi: 10.5281/zenodo.14889855. In addition, the October geophone data, also downsampled to 200 Hz over the same duration, are available.

The authors acknowledge that there are no conflicts of interest recorded.

The authors thank Ernest William Holmes for providing access to the deployment site, Kathleen Hodgkinson for her help during the October deployment, and Carly Donahue for providing guidance with the iDASv2 IU. The authors are grateful to Daniel Bowman and Elizabeth Silber for leading and providing information about the explosion series. The authors appreciate the valuable and constructive comments from three anonymous reviewers, which helped improve this article. The authors thank the Defense Advanced Research Projects Agency (DARPA) for enabling the distributed acoustic sensing (DAS) deployments during the explosion series. Both DAS deployments were supported by the Low Yield Nuclear Monitoring (LYNM) program funded by the National Nuclear Security Administration, Defense Nuclear Nonproliferation Research and Development (NNSA DNN R&D). The authors acknowledge important interdisciplinary collaboration with scientists and engineers from the Los Alamos National Laboratory (LANL), the Lawrence Livermore National Laboratory (LLNL), the Nevada National Security Site (NNSS), Pacific Northwest National Laboratory (PNNL), and Sandia National Laboratories (SNL). This article has been authored with number LA-UR‐24-33053 by Triad National Security, LLC, under contract with the U.S. Department of Energy (Number 89233218CNA000001).