Lessons from a DAS cross-well experiment at the PiTOP test site

Introduction

Distributed Acoustic Sensing (DAS) has rapidly emerged as a valuable technology for seismic monitoring and near-surface investigations, providing dense spatial sampling along fibre-optic cables installed in boreholes or at the surface. In recent years, DAS has been successfully applied to a wide range of geophysical problems, including vertical seismic profiling, cross-well imaging and monitoring of subsurface processes in both energy and environmental contexts (Re et al., 2022; Mason et al., 2025). In near-surface settings, DAS has shown particular potential for characterising shallow velocity structure, as demonstrated by recent studies using downhole and disposable fibre deployments (AlDawood et al., 2025; Stender et al., 2025).

Cross-well seismic methods are traditionally used to estimate velocity distributions between boreholes. However, their resolving power strongly depends on acquisition geometry, ray coverage, and modelling assumptions. When applied to DAS data, additional challenges arise from uncertainties in fibre-optic cable positioning, coupling conditions, and the interpretation of DAS strain measurements as equivalent seismic wavefields. These factors are particularly critical in near-surface settings, where ray paths are steep, traveltimes are short, and small geometric uncertainties may significantly affect inversion results.

This study investigates the inversion of first-arrival traveltimes extracted from a DAS cross-well dataset acquired in a near-surface environment, with an inter-well spacing of approximately 50 m. A high-frequency impulsive (sparker) source was used at multiple depths (10-100m) in a source wellbore, while a fibre-optic cable, cementing in a neighbouring borehole, recorded the seismic wavefield down to approximately 150 m depth. The objective of the study is not to obtain a highly detailed velocity image, but rather to assess the extent of velocity information that can be robustly extracted from such a dataset given its geometry and associated uncertainties. A one-dimensional layered (‘layer-cake’) inversion assuming lateral homogeneity is considered appropriate for this close inter-well spacing. Ray coverage and hit-count analyses are used to assess spatial resolution, while traveltime residuals are analysed as a function of receiver depth and time. The study aims to clarify the practical limits of DAS cross-well traveltime inversion in near-surface environments and to provide guidance for interpreting similar datasets.

Site test and acquisition geometry

The experiment was conducted at the OGS PiTOP geophysical testing site in northeastern Italy, a dedicated research facility designed to develop and validate geophysical methods under realistic field conditions (Bellezza et al., 2025; Travan et al., 2025). PiTOP hosts multiple boreholes equipped with permanent and mobile acquisition systems, providing a controlled yet versatile environment for near-surface experiments.

In this study, data were acquired using a fibre-optic Distributed Acoustic Sensing (DAS) system cemented in one borehole reaching a depth of approximately 150 m, while impulsive seismic sources were activated in a neighbouring well (Figure 1a). Seismic energy was generated by a high-frequency impulsive source (sparker) at four distinct depths: 10 m, 50 m, 75 m and 100 m. For each source depth, shots were repeated 3, 9, 10 and 8 times, respectively. The horizontal distance between the two boreholes is 47.5 m.

Dataset acquisition and processing

The DAS data were acquired with a sampling rate of 5 kHz and a gauge length of 10 m. For each source depth, individual shot records were further processed using singular value decomposition to improve the signal-to-noise ratio, prior to stacking. The enhanced records were subsequently stacked to obtain a representative response for each source depth.

The DAS system provides continuous spatial sampling along the depth dimension, enabling the extraction of first-arrival traveltimes at a large number of receiver positions. First arrivals were manually picked for each source depth, focusing on both downgoing and upgoing sections of the fibre and on the cable position considered reliable. Traveltimes were referenced to their common origin time for each shot to minimise source-related statics, as the acquisition benefited from accurate synchronisation between shot timing and DAS recording.

Figure 1b shows the result of the data processing post-stack of shots at 75 m depth in well PiTOP3 with the fibre-optic cable cemented in a looped configuration in PiTOP4. White symbols indicate manually picked first-arrival traveltimes along the fibre. The black solid line marks the end of the cable loop, corresponding to a depth of approximately 147 m. Although the fibre-optic cable is nominally vertical, small deviations from the assumed geometry cannot be excluded. Such uncertainties are expected to have a non-negligible impact on traveltime inversion, particularly in near-surface conditions where ray paths are steep and traveltimes are short.

Apparent velocities inferred from linear fits range between approximately 2280 m/s and 3235 m/s, with coefficients of determination (R2) higher than 0.93 for the source positions at 50 and 75 m (Figure 2). Linear fits were restricted to the portion of the cable below each source position in order to minimise the influence of near-surface effects and uncertainties in cable geometry. The traveltimes from the shallowest source positioned at 10 m are not considered, as the source is too close to the surface and the corresponding data presented a particularly complex wavefield propagation.

Systematic increase in apparent velocity are observed with source depth. Overall, the apparent velocities derived from the downgoing and upgoing portions of the cable are generally consistent. However, noticeable differences are observed for the deepest source position at 100 m, as well for the 75 m source, where deviations from a linear trend occur around at depths of approximately 130 m. Despite these local deviations, these apparent velocities provide a useful reference for evaluating the traveltime inversion results and are coherent with previous observations (Poletto et al., 2011).

Traveltime Inversion and Velocity Model Estimation

Figure 3 summarises the results of the straight-ray traveltime inversion applied to the DAS cross-well dataset, together with diagnostic plots of the residuals. The inversion for layer velocities was on a fixed depth grid and solved by least squares. The estimated 1D velocity model shows some depth-dependent variations, characterised by an overall increase in velocity with depth. The recovered velocities range from approximately 500 m/s to 3000 m/s. This trend is broadly consistent with apparent velocities inferred from linear fits of first-arrival moveout. However, velocity contrasts between adjacent layers remain smooth, reflecting the regularisation imposed in the inversion and the limited ray path coverage.

The residual histogram is centred close to zero, indicating the absence of a significant global bias in the traveltime predictions. Residuals are within ±10 ms, which are significant with respect to the observed first-arrival traveltimes. This magnitude indicates that modelling and/or geometry uncertainties are significant in this configuration, and that the inversion should be interpreted as a first-order estimate of vertical velocity trends rather than a high-resolution model.

Residuals plotted as a function of observed traveltime show some trends, decreasing with time and with negative value for the largest traveltimes, indicating that uncertainties accumulate along ray paths. In addition, a depth-dependent pattern is observed, with positive residuals dominating at intermediate depths and some negative residuals with a larger deviation at greater depths, probably due to a decreasing ray coverage with depth. This suggests that a purely straight-ray assumption and laterally invariant velocity layers are insufficient to fully explain the observed traveltimes. Possible contributing factors include uncertainty in the traveltime picking, possible small deviations of the fibre from a vertical trajectory and local velocity heterogeneities.

Results and discussion

Overall, the straight ray traveltime inversion provides a robust first-order velocity model that captures the main vertical velocity structure of the near-surface environment. At the same time, the residual analysis highlights the inherent limitations of a straight-ray traveltime inversion applied to this dataset in shallow near surface and close inter-well settings.

Three main depth intervals can be identified. Above 40 m, uncertainties related to the well geometry, fibre coupling effects and limited ray coverage combined with complex near-surface wave propagation are likely to affect the traveltime picks. Between 40 m and 100 m, ray coverage is optimal and velocity estimates are reliable. Slightly positive traveltime residuals in this interval may indicate an underestimation of velocity in the current model and suggest the need for minor velocity adjustments. Above 100 m, down to the end of the cable at 147 m, residuals increase in amplitude, but remain centred around zero, indicating reduced resolution without a strong systematic bias.

Apparent velocities derived from linear fits reflect integrated ray-path effects and local geometry, whereas the inversion estimates an effective smooth velocity model constrained by limited ray angles. The comparison between these two approaches highlights both the consistency of the overall velocity trends and the inherent resolution limits imposed by the acquisition geometry.

Conclusions

This study evaluates the capability of DAS cross-well first-arrival traveltime inversion for near-surface velocity characterisation using a real field dataset. For an inter-well spacing of approximately 50 m, straight-ray modelling combined with a one-dimensional layered inversion provides a robust estimate of the vertical velocity trend, consistent with apparent velocities derived from the data.

The inversion provided insights into the near-surface velocity structure, and the geometry of the wellbore where the fibre is cemented. Based on these observations, future work may benefit from improved fibre geometry characterisation, the use of curved-ray or more advanced ray-based modelling, and joint inversion strategies integrating geometry and velocity parameters to further reduce uncertainties and enhance interpretability.

Acknowledgement

The authors would like to thank all stakeholders involved in the data acquisition, processing and analysis, including the University of Padova, the Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS), Fosina and Isamgeo, for granting permission to publish these results. This project was supported by the Transnational Access funded by PNRR ECCSELLENT project (‘Development of ECCSEL - R.I. ItaLian facilities: usEr access, services and loNg-Term sustainability’).

References

• AlDawood, A., Alfataierge, E. and Shaiban, A. [2025] The Reconstruction of Robust Shallow Velocity Profile using a Disposable Fibre. Eighth EAGE Borehole Geophysics Workshop, Abstract.
• Bellezza, C., Pinna, G., Barison, E., Meneghini, F., Corubolo, P., Zgauc, F., Farina, B., Craglietto, A., Travan, A., Paganini, P., Pavan, A., Bernardi P. and Schleifer, A. [2025] Geophysical exploration case histories at the PiTOP geophysical test site-A key facility in the ECCSEL-ERIC consortium: an overview. Bulletin of Geophysics & Oceanography, 66(1).
• Mason, J., Cambois, G., Cowell, J., Kirstetter, O. and Martinez, A. [2025] Utilization of Downhole Fiber Optics to Monitor Injection Within a Carbonate Saline Aquifer Onshore Abu Dhabi. First EAGE/AAPG/ SEG Carbon Capture Utilisation and Storage Workshop, Abstract.
• Poletto, F., Petronio, L., Farina, B. and Schleifer, A. [2011] Seismic interferometry experiment in a shallow cased borehole using a seismic vibrator source. Geophysical Prospecting, 59(3), 464-476.
• Re, S., Tango, G., Del Gaudio, S. and Brovelli, A. [2022] The Role of Fibre-Optic Acoustic Sensing for the Energy Transition. First Break, 40(4), 61-67.
• Stender, V., Malehmir, A., Papadopoulou, M., Kaslilar, A. and Hamerslag, R. [2025] Collocated with Fiber-Optic Multi-Element Seismic Survey for Characterizing Mine Tailings in Central Sweden. NSG 2025 Conference on Geophysics for Mineral Exploration and Mining, Abstract.
• Travan, A., Bellezza, C., Barison, E., Corubolo, P., Meneghini, F. and Schleifer, A. [2025] Integrated geophysical studies at the OGS PiTOP testing site: an interdisciplinary approach for subsurface characterization. EGU General Assembly 2025, Abstracts.