Search for: All records

Non-invasive subsurface imaging using full waveform inversion (FWI) has the potential to fundamentally change near-surface (<30 m) site characterization by enabling the recovery of high-resolution (metre-scale) 2-D/3-D maps of subsurface elastic material properties. Yet, FWI results are quite sensitive to their starting model due to their dependence on local-search optimization techniques and inversion non-uniqueness. Starting model dependence is particularly problematic for near-surface FWI due to the complexity of the recorded seismic wavefield (e.g. dominant surface waves intermixed with body waves) and the potential for significant spatial variability over short distances. In response, convolutional neural networks (CNNs) are investigated as a potential tool for developing starting models for near-surface 2-D elastic FWI. Specifically, 100 000 subsurface models were generated to be representative of a classic near-surface geophysics problem; namely, imaging a two-layer, undulating, soil-over-bedrock interface. A CNN has been developed from these synthetic models that is capable of transforming an experimental wavefield acquired using a seismic source located at the centre of a linear array of 24 closely spaced surface sensors directly into a robust starting model for FWI. The CNN approach was able to produce 2-D starting models with seismic image misfits that were significantly less than the misfits from other common starting model approaches, and in many cases even less than the misfits obtained by FWI with inferior starting models. The ability of the CNN to generalize outside its two-layered training set was assessed using a more complex, three-layered, soil-over-bedrock formation. While the predictive ability of the CNN was slightly reduced for this more complex case, it was still able to achieve seismic image and waveform misfits that were comparable to other commonly used starting models, despite not being trained on any three-layered models. As such, CNNs show great potential as tools for rapidly developing robust, site-specific starting models for near-surface elastic FWI.

 

Active-source data acquisition included 66 vibroseis and 209 instrumented sledge hammer source locations. Multiple source impacts were recorded at each source location to enable stacking of the recorded signal. The source impacts at each source location have been aligned using cross-correlation, but to provide the most flexibility are provided unstacked (i.e., the signals from each source impact are provided separately). The active-source recordings are provided in terms of both raw, uncorrected units of counts and corrected, engineering units of meters per second. For each source impact, the force output from the vibroseis or instrumented sledge hammer was recorded and is provided in both raw counts and engineering units of kilonewtons. The passive-wavefield data includes 28 hours of ambient noise recorded over two night-time deployments. The passive-wavefield data is provided in raw counts, however, the instrument response files are provided should instrument correction be required in the future. The dataset can be used for active-source and passive-wavefield three-dimensional imaging, as well as other subsurface characterization techniques which include: horizontal-to-vertical spectral ratios, multichannel analysis of surface waves, and microtremor array measurements. 

This is a comprehensive subsurface imaging experiment in Newberry, Florida using stress waves. The site is spatially variable and contains karstic surface and underground voids and anomalies. The sensing technologies used comprised a dense 2D array of 1920 DAS channels and a 12 x 12 grid of 144 SmartSolo 3C nodal stations, which covered an area of 155 m x 75 m and were used to record both active-source and passive-wavefield data. The active-source data was generated by a variety of vibrational and impact sources, namely: a powerful three-dimensional vibroseis shaker truck, a 40-kg propelled energy generator (PEG-40kg), and an 8-lb sledgehammer. The vibroseis shaker truck was used to vibrate the ground in the three directions at 260 locations inside and outside the instrumented area, while the impact sources were used at 268 locations inside the instrumented area. In addition to active source data, four hours of ambient noise were recorded using the DAS, while the nodal stations recorded 48 hours of ambient noise in four 12-hour increments over a period of four days. The waveforms obtained from the 1920 DAS channels for every active-source shot or passive-wavefield time block were extracted, processed, and stored in H5 files. These files can be easily visualized using a Python script incorporated with the open-access dataset. Additionally, the three-component data gathered from each SmartSolo nodal station were consolidated into a single miniSEED file, and the data from all 144 nodal stations obtained during each active-source shot or passive-wavefield time block were extracted and saved into a separate folder. We anticipate that this dataset will be a valuable resource for researchers developing techniques for void and anomaly detection using noninvasive, stress wave-based subsurface imaging techniques. 

Abstract Non-invasive surface wave methods are increasingly being used as the primary technique for estimating a site’s small-strain shear wave velocity (Vs). Yet, in comparison to invasive methods, non-invasive surface wave methods suffer from highly variable standards of practice, with each company/group/analyst estimating surface wave dispersion data, quantifying its uncertainty (or ignoring it in many cases), and performing inversions to obtain Vs profiles in their own unique manner. In response, this work presents a well-documented, production-tested, and easy-to-adopt workflow for developing estimates of experimental surface wave dispersion data with robust measures of uncertainty. This is a key step required for propagating dispersion uncertainty forward into the estimates of Vs derived from inversion. The paper focuses on the two most common applications of surface wave testing: the first, where only active-source testing has been performed, and the second, where both active-source and passive-wavefield testing has been performed. In both cases, clear guidance is provided on the steps to transform experimentally acquired waveforms into estimates of the site’s surface wave dispersion data and quantify its uncertainty. In particular, changes to surface wave data acquisition and processing are shown to affect the resulting experimental dispersion data, thereby highlighting their importance when quantifying uncertainty. In addition, this work is accompanied by an open-source Python package, swprocess , and associated Jupyter workflows to enable the reader to easily adopt the recommendations presented herein. It is hoped that these recommendations will lead to further discussions about developing standards of practice for surface wave data acquisition, processing, and inversion. 

Quantitative dynamic strain measurements of the ground would be useful for engineering scale problems such as monitoring for natural hazards, soil-structure interaction studies, and non-invasive site investigation using full waveform inversion (FWI). Distributed acoustic sensing (DAS), a promising technology for these purposes, needs to be better understood in terms of its directional sensitivity, spatial position, and amplitude for application to engineering-scale problems. This study investigates whether the physical measurements made using DAS are consistent with the theoretical transfer function, reception patterns, and experimental measurements of ground strain made by geophones. Results show that DAS and geophone measurements are consistent in both phase and amplitude for broadband (10 s of Hz), high amplitude (10 s of microstrain), and complex wavefields originating from different positions around the array when: (1) the DAS channels and geophone locations are properly aligned, (2) the DAS cable provides good deformation coupling to the internal optical fiber, (3) the cable is coupled to the ground through direct burial and compaction, and (4) laser frequency drift is mitigated in the DAS measurements. The transfer function of DAS arrays is presented considering the gauge length, pulse shape, and cable design. The theoretical relationship between DAS-measured and pointwise strain for vertical and horizontal active sources is introduced using 3D elastic finite-difference simulations. The implications of using DAS strain measurements are discussed including directionality and magnitude differences between the actual and DAS-measured strain fields. Estimating measurement quality based on the wavelength-to-gauge length ratio for field data is demonstrated. A method for spatially aligning the DAS channels with the geophone locations at tolerances less than the spatial resolution of a DAS system is proposed.