Spatial Characterization of Shallow Structures in the Revell Batholith Integrating Seismic Imaging Techniques

Abstract

The Revell Site, located in Northwestern Ontario within the Canadian Shield, is being assessed as a potential Deep Geological Repository (DGR) for Canada’s used nuclear fuel. Effective DGR establishment requires comprehensive subsurface assessment, particularly in evaluating structural stability, hydrogeological attributes, geological composition, and geochemical properties. Key among these considerations is understanding the three-dimensional characteristics of structural features to ensure the site’s suitability for long-term containment of radioactive materials. This case study focuses on imaging and characterizing structures within the predominantly intact biotite granodiorite-tonalite host rock at the Revell Site. Borehole data reveals these structures as mostly sub-horizontal mafic intrusions, with thicknesses of up to 3.5 m. They appear either as separate, discrete entities or in stacked configurations. Despite their limited thickness, most of these intrusions exhibit discernible attributes in surface seismic images due to their distinct physical properties. Some, however, exhibit velocity variations that decrease their overall reflectivity properties. An integrative approach employing surface seismic, Vertical Seismic Profiles (VSP), and borehole data facilitates the spatial identification of over 30 of these mafic structures. This comprehensive characterization not only lays the foundational framework for future discrete fracture network models but also provides important support for simulations related to fluid flow dynamics, groundwater behavior, contaminant dispersion, and heat transport mechanisms within the Revell Site. This study underscores the critical role of in-depth subsurface characterization in ensuring the secure, long-term management of radioactive materials in DGRs.

1 Introduction

Radioactive waste generated during nuclear power production, research, and other nuclear applications requires careful handling, storage, and disposal to ensure the short- and long-term safety of both humans and the environment. Deep geological repositories are widely accepted internationally as the preferred solution for long-term storage of high-level radioactive nuclear waste, such as used nuclear fuel. Countries like Finland, Sweden, France, Switzerland, and Canada are progressing through site selection, characterization, and infrastructure development for safe storage solutions. As Canada proceeds with its site selection process, a range of important geological factors are being evaluated. These include develo** a thorough comprehension of the geological setting, assessing geotechnical suitability, analyzing hydrogeological conditions, ensuring seismic stability, and addressing long-term safety considerations (NWMO, 2022). Seismic data play a vital role in the site characterization process by enabling the imaging of subsurface architecture and providing quantitative interpretation of geosphere properties. The effectiveness of both surface reflection seismic and vertical seismic profiles (VSP) for imaging complex geological environments in crystalline terrains has been demonstrated (Green & Mair, 1983; Greenwood et al., 2012; Villamizar et al., 2022). The integration of these two different seismic techniques has significantly improved the reliability of structural interpretation across various geological settings (Hurich & Deemer, 2013; Martínez & Mendoza, 2011) and has been extensively utilized for characterizing potential sites for nuclear waste disposal, as evidenced by the work of Cosma et al. (2003), Juhlin and Stephens (2006), and Li et al. (2021).

By employing appropriate acquisition geometries, data-driven processing strategies, and a collaborative seismic interpretation approach, we effectively characterize the occurrence, geometry, and spatial distribution of structures that are on the resolvable limit of reflection seismic using two-dimensional (2D) surface seismic and walkaround VSP data at a site, hereafter referred to as the ‘Revell Site’, underlain by Archean-aged granitoid bedrock of the Revell batholith in the Canadian Shield. Based on our integrative approach that incorporates 6000 m of borehole information, including geophysical logs, geochemical analysis, lithological descriptions, and core samples we demonstrate that many of these structures correlate with a suite of gently dip** mafic dykes.

These mafic dykes are of particular interest because their occurrences, although volumetrically minor, are often associated with increases in fracture intensity, enhanced strain localization (particularly localized along the dyke margins), quartz and calcite veining, and higher alteration intensities within the otherwise sparsely fractured biotite granodiorite-tonalite host rock. Decimetre- to metre-scale hematite ± sericitization and chloritization haloes commonly envelop the dykes. Water-conducting fractures are sometimes present within, or along the contacts of these mafic dykes. These dykes are characterized by their discrete nature and limited thickness, typically 3.5 m or less. Despite their limited thickness, most of the intrusions exhibit notable physical properties such as high bulk density and high velocity relative to the surrounding rock, which make them detectable in seismic surveys. By studying the distinct seismic response of these dykes, we effectively delineate their presence in areas away from borehole locations. These intrusions are observed to be dispersed throughout the entire Revell Site, with higher concentrations within specific rock domains. Initial core observations and petrographic analysis defined these mafic dykes as amphibolites. However, recent rock chemistry results suggest a possible lamprophyre classification. Despite this uncertainty in classification, for simplicity and consistency, henceforth these units are collectively referred to as mafic dykes. Whether they are lamprophyre or amphibolite intrusions, their seismic response is mainly influenced by their mafic composition relative to the surrounding felsic rock.

Understanding the characteristics of these intrusions within the geological framework of the Revell Site is important for identifying structures that may serve as potential pathways for the transport of radionuclides to the biosphere. Knowledge of these characteristics is crucial for supporting the development of discrete fracture network (DFN) models and other simulations related to fluid flow such as groundwater, contaminant, and heat transport modeling and forecasts. By incorporating insights on the mafic dykes, our understanding of the potential migration routes and mechanisms through which radionuclides may be transported is enhanced, thus contributing to overall geological site understanding and supporting future safety assessments, should the Revell Site be selected to host a DGR.

We start this case study by providing an overview of the regional and local geology surrounding the Revell Site, situated within the Revell batholith that was emplaced approximately 2.7 billion years ago (NWMO, 2022). This overview includes a review of the available borehole data obtained from six boreholes, each approximately 1000 m in length. Subsequently, we provide a brief description of the acquisition and processing techniques employed for the six 2D crooked seismic lines and three walkaround VSPs used for subsurface imaging. The final section of this paper is dedicated to the seismic interpretation of mafic dykes, where we employ 1D forward modeling to identify their distinct signature on surface seismic data. We further integrate and interpret multiple data sources, including borehole lithological logging and migrated VSP sections. By employing this approach, we describe the geometry and position of over 30 mafic dykes throughout the Revell Site. Our results are consistent with the observations of these dykes in the available borehole data.

2 Geological Setting of the Revell Site

2.1 Regional Geology

The ca. 2.70–2.67 Ga Revell batholith is south-centrally located in the western Wabigoon terrane in the northwestern part of the Superior Province of the Canadian Shield (Stott, 2011; Stott et al., 2010). Easterly trending lithotectonic terranes subdivide the Superior Province into tectonically bounded domains with characteristics distinct from adjacent regions prior to their accretion into the Superior Province (Percival & Easton, 2007; Stott et al., 2010). The Revell batholith is approximately 10 km by 40 km in size, elongated in a northwest-southeast orientation, and abuts the Marmion terrane to the south. Szewczyk and West (1976) modelled the Revell batholith to be a tabular intrusion with an approximate thickness of 1.6 km. Recently acquired geophysical surveys used in three-dimensional (3D) coupled gravity and magnetic modelling suggest that the batholith is on the order of 4 km deep with a relatively flat base (Mushayandebvu et al., 2023).

Archean supracrustal rocks belonging to the Bending Lake and the Raleigh Lake greenstone belts wrap entirely around the margins of the Revell batholith (Fig. 1) and are affected, to varying degrees, by penetrative brittle-ductile to ductile deformation under greenschist- to amphibolite-facies metamorphic conditions (Blackburn & Hinz, 1996). Uranium–lead (U–Pb) geochronological analysis of these supracrustal rocks produced ages that range between 2734.6 ± 1.1 and 2725 ± 5 Ma (Stone et al., 2010).

Fig. 1

a Geological terrane subdivision of the Superior Province in Northwestern Ontario (after Stott, 2011). b Northern part of the Revell batholith and surrounding mafic and intermediate metavolcanic rocks of the Raleigh Lake and Bending Lake greenstone belts. The Revell Site, shown in the black box, is situated centrally in the northern part of the batholith. Red diamonds and white lines show borehole locations and trajectories

The extensively studied Revell batholith (Parmenter et al., 2020; and references therein), now under consideration as a potential DGR site (NWMO, 2022), comprises four primary rock suites (Fig. 1). The oldest phase consists of white to grey, medium-grained biotite tonalite to granodiorite, with an age of approximately 2734.2 ± 0.8 Ma. A slightly younger phase is characterized by grey to white, hornblende tonalite to granodiorite (Stone et al., 2010). The central, northern, and southern regions are dominated by light grey to white, medium-grained biotite granite to granodiorite with U–Pb ages between 2700 ± 4 and 2712 ± 5 Ma based on dating of 4 drill core samples (Davis, 2022). The youngest phase is a potassium (K)-feldspar megacrystic granite with an age of 2694 ± 0.9 Ma. Additionally, there are Proterozoic-aged diabase dykes (Wabigoon dyke dwarm) extending across the batholith, ranging from approximately 15–30 m in width (Lenauer et al., 2017).

2.2 Conceptual Understanding of the Site

2.2.1 Bedrock Geology

The Revell Site is entirely underlain by rocks assigned to the biotite granite to granodiorite suite. Detailed map** campaigns determined that the bedrock in this area ranges compositionally between granodiorite and tonalite, with gradational contacts masking their intrusive relationship, and lesser granite that intrudes both compositional endmembers (Lenauer et al., 2017). A structural lineament interpretation was completed over the site and surrounding area in order to begin to characterize the fracture network (DesRoches et al., 2018). Between 2017 and 2021, six boreholes were drilled, cored and logged to characterize the subsurface geology beneath the site. Rock type, alteration, and structural properties were all logged from approximately 6000 m of continuous HQ-size core. During logging, core samples were collected for laboratory analysis of geomechanical, petrophysical, and porewater characteristics. IG_BH01 was drilled nearly vertical to evaluate the integrity of the bedrock away from fracture zones. Boreholes IG_BH02 through IG_BH06 were inclined approximately 70° with specific azimuths chosen to increase the likelihood of intersecting steep fractures and inferred fracture zones in the subsurface.

Consistent with map** observations at the surface, the dominant rock type in recovered core is a grey, medium-grained biotite granodiorite-tonalite to biotite tonalite representing more than 95% of the bedrock encountered. The main rock type is punctuated by felsic dyke occurrences that are compositionally similar to the granodioritic to tonalitic bedrock, and by the suite of mafic dykes which are a focus of the current study. The mafic dykes are dark grey-green, equigranular, granoblastic and usually fine-grained. They are also compositionally distinct from the surrounding felsic rocks, as evidenced by their SiO₂ (41–55 wt%), Fe₂O₃ (7–11 wt%), TiO₂ (0.5–0.9 wt%), and MgO numbers (5–17 wt %), and enrichment in Cr (190–1460 ppm), Co (26–62 ppm), and Ni (75–535 ppm) (Fig. 2). Radiometric U–Pb dating of a single mafic dyke yielded an average ²⁰⁷Pb/²⁰⁶Pb age of ca. 2700 ± 4 Ma (Davis, 2022). The zircon grains exhibit a slightly rounded, possibly resorbed, appearance suggesting they are xenocrystic. This interpretation provides a maximum emplacement age and further suggests a co-magmatic intrusive association with the surrounding granodiorite-tonalite bedrock, which yielded ²⁰⁷Pb/²⁰⁶Pb ages ranging from 2712 ± 2 to 2700 ± 4 Ma (Davis, 2022).

Fig. 2

Cross-plot results showing clear geochemical differentiation between the mafic intrusives (cross symbol outlined in an oval), felsic dykes and the main biotite granodiorite-tonalite rock types (circle symbol outlined in a box). Cross-plots show SiO₂ versus a Cr, b Co, c MgO, and d Fe₂O₃

The mafic dykes show evidence of penetrative ductile to brittle-ductile deformation obscuring any indication of their primary relationship with the surrounding bedrock. Deformation manifests as cm- to dm-scale localized shear zones that are developed sub-parallel to their contacts. These dykes, where logged in core, have a mean orientation of 24° towards the northwest (348°). Orientations were measured using optical and acoustic televiewer logs and characteristics were recorded from drill core. Figure 3 shows the stereonet of poles to dyke planes of all six boreholes.

Fig. 3

Stereonet of poles to mafic dykes (red circles) logged in six core boreholes. Mean orientation pole (black square) of the mafic dykes plots in the SE quadrant with a great circle dip** 24° towards the north-northwest. Pole density contouring using a fisher distribution with a 5% count circle area

2.2.2 Geophysical Log Response

Continuous in-situ measurements of P-wave velocity (V_p), S-wave velocity (V_s), and bulk density (ρ) were logged in each borehole. Elastic moduli and acoustic impedance were derived for each borehole using the measured V_p, V_s, and ρ logs. Normal resistivity, magnetic susceptibility, natural and spectral gamma, neutron, and optical and acoustic televiewer logs were recorded to aid in distinguishing contacts and trends in rock type, mineralogy, and alteration as well as map** structural orientations.

Mafic dykes consistently exhibit elevated logged bulk density, often coupled with increases in velocity, relative to the host rock. Figure 4 shows geophysical logs along IG_BH01 depicting ρ, V_p, acoustic impedance (AI), and fracture intensity with color-coded rock types. Mafic dyke intervals (I1–I8) are also included in this figure. These intervals do not correspond to specific type of dykes. They are merely a nomenclature used in this paper to facilitate the discussion and draw conclusions more conveniently regarding their seismic signatures. Based on borehole data, on average, the mafic dykes are characterized by an elevated P-wave velocity with a median value of 5882 m/s, compared to 5660 m/s for the main biotite granodiorite tonalite. Similarly, mafic dykes have higher median bulk rock density of 2.8 g/cm³ compared to 2.66 g/cm³ for the main bedrock. As a result, derived median acoustic impedance for all mafic dykes is roughly 7.5% higher than the main biotite granodiorite tonalite bedrock. Locally, however, some mafic dykes present higher bulk density than others, and can be coupled with either an increase or reduction in P-wave velocity. This variable response tends to correlate with both the thickness or proximity of clustering dykes, which empirically increases the P-wave velocity and density, or with an increase in fracture intensity that reduces P-wave velocity. The nature of this response is elaborated further in Sect. 4.2. We believe seismically visible mafic dykes are typically those with densities and velocities higher than the median values derived from borehole data.

Fig. 4

Bulk density, p-wave velocity, acoustic impedance and fracture intensity (per 1 m) for the near vertical borehole IG_BH01. Due to the vertical scale of 1:10,000, some thin dykes may not be visible in rock type log

2.2.3 Fracture Characteristics

Core and televiewer logs are used to assess the site-scale network of fractures (i.e., joints, veins, and faults) and their characteristics, including orientations, geologic apertures, mineral infills, and intensities along the borehole. Fracture intensities are calculated per meter along the borehole length and are used to define specific zones of elevated intensity. Across the entire site, the upper portion of the bedrock (~ 100 m) tends to have an elevated fracture intensity and the highest presence of moderately altered rock.

Based on the compilation of all borehole structural information, four main fracture sets are defined for the site, including one sub-horizontal fracture set and three steeply-dip** fracture sets striking northwest, north-northeast, and east-northeast. Similarly-oriented fracture sets are identified during surface map**. Isolated, quartz-filled, ductile shear zones (~ 1–30 cm thickness) are similarly oriented to the sub-horizontal and northwest-striking fracture sets, suggesting late- to post-magmatic re-activation of joints developed during cooling of the intrusion.

Overall, there is a gently north-dip** structural grain characterized by both ductile and brittle structures with a mean dip of 18° towards 354°, similar to the mean mafic dyke orientations (see Fig. 3). Locally, sub-horizontal fractures are clustered in discrete intervals of the borehole with intensities up to 30 per meter. Many of these intervals correlate with mapped mafic dykes, or other tonalitic subordinate rock types, with orientations generally subparallel to sheared dyke contacts.

Steeply dip** fractures mapped in boreholes are locally clustered, and in some instances these fracture clusters can be correlated to interpreted lineaments projected into the subsurface. In borehole, these fracture cluster intervals are characterized by elevated fracture intensities, with quartz, epidote, chlorite, and hematite mineral infill. Intervals with steeply-dip** fracture clusters can range from approximately 10–25 m length along borehole, with peak fracture intensities of approximately 20 to 30 fractures per meter. These intervals also tend to show presence of weak to moderate alteration intensity.

A general characteristic of the Revell Site is that presence of mafic dyke occurrences tends to be associated with observed increases in borehole fracture intensities, with overall decreases in intensity with distance away from the mafic dyke. A distinct peak in fracturing is evident within and adjacent to these dykes, in contrast to a relatively steady lower fracture intensity at distances beyond approximately 25 m from any single occurrence as indicated by the truncated power-law distribution fit to the normalized probability density (Fig. 5). The r² for the fit with the truncated power-law distribution is 0.998. Importantly, borehole hydraulic testing indicates that sub-horizontal fractures are more commonly water-conducting, especially where associated with these mafic dykes, or other tonalitic subordinate rocks. These relationships reinforce the importance of develo** a sufficient understanding of the spatial distribution of mafic dykes through the site, since the tonalitic subordinate rocks are seismically transparent.

Fig. 5

Normalized probability density of fractures in proximity to mafic dykes in six deep boreholes. The dashed line represents a fit with truncated power-law distribution

3 Data Acquisition and Processing

3.1 Seismic Data

The primary objective of the 2D seismic surveying campaign was to improve the understanding of the upper 1500 m of the subsurface at the Revell Site, with a focus on correlating seismic reflection data with lithological and structural information derived from borehole data. We wanted to image potential geological features, both sub-horizontal and gently dip** (with inclinations less than 45°), that may produce recognizable seismic reflections at depths ranging from 300 to 700 m.b.g.s (meters below ground surface), up to a maximum depth of 1500 m.b.g.s. Likewise, we aimed at detecting, tracing, and analyzing seismic reflections in the data that may correspond to fracture zones or lithological boundaries.

3.1.1 Acquisition of Surface Seismic

A set of high-resolution 2D seismic reflection lines were acquired in the Northern section of the Revell Batholith between November and December of 2020. Optimal survey configurations that enhance the imaging success and accuracy of the geological features of interest were designed. However, because the terrain at the study area is not easily traversable, seismic operations relied on existing roads and bush tracks. This resulted in the acquisition of non-straight seismic transects with different degrees of crookedness. A total of six 2D seismic transects were collected as displayed in Fig. 6. Note that Line 1 is divided into two lines, Line 1.1 and Line 1.2, during processing.

Fig. 6

a Location of seismic acquisition lines (black) and stacking lines (blue), and b location of vertical seismic profile shot points (triangles). VSP surveys were acquired in IG_BH04, IG_BH05 and IG_BH06. Shot locations for VSP survey in IG_BH06 are green triangles. Shot point 82 is discussed in Sect. 4.1.2. Red diamonds show borehole collar locations, with a white line showing the trajectory. Background shows hill-shaded topography. Geodetic datum: NAD83 UTM15N

Most of the seismic transects, albeit crooked, follow either a northwest or northeast shooting orientation to maximize the imaging of dip-line structures known to exist at the Revell Site. We use the crookedness of the profiles to our advantage as off-plane reflectors could also be imaged by adequately processing the swath distribution of midpoint traces, which covers a larger subsurface area compared to surveys conducted along straight-line geometries. The sinuosity index of these lines, which is the ratio of the total distance of the original (crooked) profile to the distance of a straight-best-fitting line, is computed. This results in an average sinuosity index of approximately 1.15, with the highest index being Line 1 (1.28), followed by Line 4 (1.18). The sinuosity of these seismic lines allows the imaging of cross-dip structures not necessarily within the vertical plane below the stacking line.

The acquisition parameters are displayed in Table 1. A vibroseis VIBSIST-3000 system and single-array 10-Hz SM-24 geophones were employed for the acquisition of all transects during the seismic survey. At least 200 receivers were live for every shot, guaranteeing a minimum coverage of 2000 m per source station. The number of receiver and source stations varied per line, depending on its length. For instance, Line 1 utilized the highest number of stations, comprising 231 source points and 560 receiver stations for a total receiver length of 5329 m. All receiver lines were extended by an additional 500 m into the forested areas to enhance effective subsurface imaging. On average, the interval between source points was 20 m, while that of receiver stations was 10 m.

Table 1 Surface seismic acquisition parameters

In Fig. 7, a representative trace-normalized vibroseis gather from Line 3 is presented to 600 ms, 40% of the recording length. The gather exhibits minimal noise levels prior to the onset of the first arrivals, and the spread covers around 2800 m. Clear refracted energy is observed up to a maximum offset of roughly 1500 m. The signal-to-noise ratio appears to be favorable at both near and far offsets. However, there is noticeable contamination from surface waves and converted modes, indicated by the blue arrows. These arrivals obscure the reflected energy, making it difficult to identify in the shot gather. Figure 7 also depicts the frequency-wavenumber spectrum of the gather, highlighting the prevalence of surface arrivals and wave conversions in the low-frequency portion of the spectrum.

Fig. 7

Representative trace-normalized shot gather and its corresponding frequency-wavenumber spectrum obtained from Line 3

3.2 Processing of Surface Seismic

Two-dimensional seismic data processing in crystalline environments is hindered by their commonly crooked survey geometry. Non-straight seismic transects in non-layered media violate the very foundations of the conventional common-midpoint (CMP) processing approach, which assumes physical properties only vary along the in-line direction (Nedimović & West, 2003; Villamizar et al., 2022). In cases where straight-line acquisitions and a cylindrical Earth cannot be assumed, more comprehensive processing techniques must be implemented. We perform thorough static corrections and implement three-dimensional stacking and migration techniques on the crooked seismic data, adequately accounting for reflection points scattered perpendicular to the shooting line. Below, we present a summary of the processing steps applied to the six seismic sections.

Over 17 km of 2D crooked seismic reflection data are processed following the steps described in Table 2. The seismic records obtained through the Swept Impact Seismic Technique (SIST) are decoded and shot gathers are assigned its corresponding geometry, followed by a zero-phase band-pass frequency filter with corners 20 and 130 Hz. These corners are selected upon evaluating the amplitude spectrum of all recorded shot gathers using different frequency bands prior to first break picking. Surface waves and converted modes pose a challenge due to their high amplitudes and low-frequency characteristics, which result in interference.

Table 2 Seismic reflection processing sequence

First break picking of P waves is reliably performed on most records as seismic noise levels are low prior to the onset of the first arrivals. The obtained traveltime picks are utilized to construct a 3D tomographic model of the shallow subsurface to assist in the implementation of static corrections; we set the final datum for all lines at 450 m.a.s.l (meters above sea level). A best-fitting line method is applied to each seismic transect to extract a slice of the velocity model along their crooked geometry. This is because the computation of velocity is performed in 3D space, properly accounting for the crookedness of the profiles. Note that Line 1 is divided into two sections during processing because its degree of crookedness would not have rendered enough coverage for a three-dimensional traveltime inversion. Therefore, two different stacking lines are used along the original geometry following a linear best-fit approximation as depicted in Fig. 6 by Line 1.1 and Line 1.2. Upon examining the resulting velocity model (not included here), we find that the transition from the more fractured and weathered near-surface zone to competent bedrock occurs at an average depth of 75 m for all the surveyed lines.

The removal of P-wave first arrivals and the suppression of surface waves and converted modes is achieved by editing/muting traces and implementing variable slope median filters. We implement 3D pre-stack Kirchhoff depth migration using the best-fit 2D lines shown in Fig. 6. Similar to the traveltime tomography implementation discussed above, these lines are utilized solely as projection axes as the computation of migration paths and reflection points is performed in 3D. We recognize that the complexity of near-surface imaging is further influenced by out-of-plane energy, which arises from the crookedness of the profiles and their oblique orientation in relation to the main structural features. The combined influence of these factors leads to considerable contamination in the first 300 m of the subsurface. This contamination is evident in the post-processed reflectivity images, as demonstrated in Sect. 4.1.1.

3.2.1 Surface Reflectivity Images on Crooked Profiles

Subsurface imaging using seismic reflection techniques in hard rock environments is more complex than in sedimentary settings partly because of the various phases of deformation and alteration that the crystalline rocks have undergone during their history. Likewise, the suitability of conventional seismic reflection processing is compromised by inadequate acquisition geometries in addition to geological variations and the geometric arrangement of subsurface reflectors. Seismic wavefields are more prone to scattering in such cases, where undesired arrivals such as surface waves, converted modes, and seismic multiples dominate seismic records (Beckel & Juhlin, 2019; Malehmir et al., 2012; Snyder et al., 2009; Villamizar, 2021).

Seismic reflectivity imaging using the 2D crooked geometries at the Revell Site are subjected to challenging aspects such as the geometrical relationship between CMP bins and seismic-scale reflectors/scatterers, oblique shooting line direction and preferential geometry of target reflectors, near-surface out-of-plane back scattering, and surface wave dominance in shallow and fractured media. These factors can prevent proper energy focusing during stacking. Cross-profile energy can be stacked in-phase and incorrectly migrated as in-line reflections, leading to cross-profile events present within the vertical plane containing the two-dimensional processing line. Therefore, the potential for out-of-plane reflections and in-line reflections to appear smeared as a result of acquisition limitations and processing artifacts further complicate our interpretation, leading to potential subjectivity. We adopt an integrative interpretation approach to manage these limitations. This involves jointly interpreting multiple seismic lines, some of which intersect each other at nearly right angles, integrating borehole geophysical data, and analyzing multi-borehole VSPs. By combining the analysis of multiple data sources, we improve our confidence in the identified structural features.

3.2.2 Acquisition of Vertical Seismic Profiles

Vertical Seismic Profiles were acquired in three of the six boreholes displayed in Fig. 6. Boreholes IG_BH04, IG_BH05, and IG_BH06 are used to collect multi-azimuth and multi-offset VSP data. The main motivation for VSP data acquisition is to increase the imaging success of geological structures with both steep and gentle dips around the boreholes, as well as to understand their spatial distribution, supplementing the data obtained from surface seismic surveys. Because the collected VSP data have wide arrivals in most cases, they provide much closer illumination to the surface seismic profiles away from borehole tracks, allowing for cross-validation and a joint structural interpretation.

Due to limited accessibility of the terrain, the acquisition geometry of all VSPs can be more accurately described as modified walkaround VSP geometries. A representative example of such a geometry is shown by the shot points of the VSP collected in borehole IG_BH06 (green triangles on Fig. 6). These shot points almost surround the deviated borehole IG_BH06, allowing for more azimuthal coverage. The black triangles in Fig. 6 depict the position of additional shot points used by the other two boreholes. The location of some of these shot points is common to all three boreholes.

A summary of the VSP acquisition parameters for the three boreholes is listed in Table 3. The seismic source is the same type of vibroseis source used to collect the surface seismic data, whereas the receivers consist of an array of 12 levels 3-component RD-XYZH geophones spaced at 5-m intervals over the depth range 80 m and 980 m along each borehole. This receiver spacing guarantees the discrete reconstruction of the continuous seismic wavefield, spatially unaliased, for the survey’s maximum frequency and minimum propagating wavelength in accordance with the Nyquist criterion (Vermeer, 1990). Under this criterion, the receiver sampling must be less or equal to half the minimum propagating wavelength, which for all three boreholes is approximately 11 m.

Table 3 Acquisition parameters of VSPs

A total of 30 shot stations are used for each borehole, with approximately 180 receiver stations covering 900 m at 5 m intervals. The shot stations have varying offsets, which refer to the distances between the top of the borehole and each source. Specifically, for IG_BH04, these offsets span from 24 to 1035 m. In the case of IG_BH05, they range from 25 to 1017 m, and for IG_BH06, they extend from 32 to 906 m. The acquisition method involved initially acquiring all shot points for each 12-level receiver array. This means that the 12-level geophone string was kept clamped to the borehole at the same depth interval. To ensure the quality of the coupling mechanism, the last record of each day was repeated at the beginning of the subsequent day.

3.2.3 Processing of Vertical Seismic Profiles

We follow a conventional VSP processing approach similar to that employed by Sullivan et al. (2014) and Takam Takougang et al. (2019). The processing steps are summarized in Table 4. We examine every intermediate step for data quality and identification of possible sources of noise, extensively editing traces with high noise. Our processing involves the time stack of the VIBSIST impact sequences followed by trace selection, sorting, and final geometry assignment. The retained overall frequency content of all shot points lies between 20 and 250 Hz, with significant contamination of surface waves in the frequency bands lower than 50–100 Hz. We implement a zero-phase band-pass frequency filter, spectrally equalizing the VSP data from 50 to 250 Hz. This further assists in mitigating the interference of undesired secondary arrivals.

Table 4 Processing parameters of VSPs

Assuming a source-receiver line polarization of the P-wave, we computationally perform the rotation of the horizontal components into the radial (in line with the direction of the incident P wave), transverse (out of line with the direction of the incident P wave), and axial (directed along the borehole) coordinate system for each of the shot positions. This is necessary as the orientation of the transverse components is not set during acquisition. Traveltime picking is performed on all shot gathers from the rotated profiles and velocity fields are derived from tomographic reconstruction using the first-break traveltimes. We present an example demonstrating the correspondence between the P-wave and S-wave profiles obtained from the tomographic velocity model and those obtained from the geophysical full-waveform sonic log along the borehole IG_BH05 trace. Although the resulting tomographic inversion is not directly related to the structures interpreted on the processed and migrated VSP profiles, it does establish indirect reliability of the final structural interpretation discussed later in this case study.

The velocity profiles for both P-wave and S-wave are displayed in Fig. 8 starting from a depth of 70 m and extending to approximately 1000 m. The fracture frequency curve illustrates the number of fractures identified at 5-m intervals along the depth axis and serves as a reference for fracture occurrence. Significant drops in both P- and S-wave velocities are predominantly observed in areas with high fracture density along the logged curves. One notable example is the peak observed at approximately 800 m depth, which indicates the presence of relatively extensive fracturing in that region. One can observe that the sensitivity of the tomographic reconstruction to local velocity variations is lower compared to the velocities obtained from well logging, as expected. The inversion of S-wave velocity is exceptionally accurate, whereas a slight overestimation of P-wave velocities can be seen in the depth range of approximately 300–1000 m. However, the average velocity from the P-wave well log is 5800 m/s and that derived from tomography is 5950 m/s. This represents a difference of less than 3%, which is well within the resolvable limit of our VSP data. The variation in wavelengths between the two datasets is less than a meter at the highest sampling frequency. Hence, the observed velocity difference depicted in Fig. 8 does not affect the reliability of the tomographic reconstruction.

Fig. 8

Fracture frequency per 5 m and velocity profiles from borehole logging (dashed lines) and VSP tomographic inversion (solid lines) along IG_BH05

After removing direct P- and S-wave arrivals, applying amplitude compensation and equalization, and implementing static corrections, we migrate all the VSP data following the 3D image point migration approach of Cosma et al. (2010). Section 4.1.2 showcases examples of the processed and migrated images, along with their corresponding interpretation.

4 Results and Discussion

4.1 Seismic Profile Observations

4.1.1 Surface Seismic

As observed in Fig. 7, the seismic records are dominated by surface waves with strong ground-roll presence as well as the contribution from shallow out-of-plane energy. Despite our efforts to reduce the impact of these arrivals by implementing noise attenuation techniques during reflection processing, remnant ground roll still contaminates the final stacked and migrated sections. Therefore, we apply post-stack L2-norm data normalization to all the migrated seismic sections prior to the qualitative interpretation of seismic reflections. This step is necessary as coherent and continuous reflections are traced at depth only after equalizing seismic amplitudes due to the disproportionate contribution of near-surface noise. Note, however, that the interpretation process is cross validated using the non-normalized data. This is done to ensure that no interpretation artifacts are introduced as a result of visual bias stemming from seismic equalization. Figure 9a displays the original non-normalized seismic Line 1.2 and Fig. 9b displays the same transect after L2-trace normalization. Figure 9a displays significant presence of coherently stacked arrivals caused by surface wave contamination. These arrivals are apparent within the elevation range from surface to approximately 50 m.a.s.l, and they tend to obscure deeper reflections. This effect is particularly pronounced along the flanks of the profile. In contrast, reflections at depth are more distinct in Fig. 9b, which ease their traceability.

Fig. 9

Representative seismic transect (Line 1.2) a before and b after L2-trace normalization

Most of the 2D crooked transects exhibit a widespread presence of strong and, albeit discreet in nature, coherent reflections, which is particularly apparent in the elevation range between 50 and − 850 m.a.s.l. The continuity of these reflections varies per individual line, ranging from a few tens of meters to several hundred meters. In some cases, they reach kilometer-lengths as some reflections are consistently identified in multiple lines. This spatial relationship among reflections can be appreciated when several adjacent lines are analyzed together as demonstrated in subsequent sections. The most prominent seismic reflections, in terms of continuity and energy contrast, are sub-horizontal to gently dip** in all crooked profiles, correlating with the structural fabric highlighted above in Sect. 2.2.3.

Our structural interpretation is primarily focused on characterizing the 3D expression of these gently dip** reflections, which are related to mafic dykes concentrated between 50 and − 850 m.a.s.l. The identification of these dykes in both surface seismic and VSP profiles is supported by borehole data and demonstrated in Sect. 4.2.3. To mitigate the impact of acquisition and processing constraints discussed above and reduce interpretation biases, the geometry of mafic dykes is inferred by analyzing prominent and continuous reflections in intersecting seismic lines. This approach enables the estimation of the true dip, orientation, and three-dimensional extent of these reflections. For reflections that are restricted to individual seismic lines, on the other hand, their geometry is determined through a combination of VSP interpretations.

It should be noted that although we have identified several other prominent and coherent reflections in the surface seismic profiles, their nature could not be directly associated with mafic dykes due to insufficient multi-source data. However, given their similar seismic expression, such as distinct amplitude contrast, apparent dip, and continuity (as illustrated in Sect. 4.2.2), it is probable that these reflections correspond to the same type of intrusions. These characteristics make them an important feature to characterize should a deep geological repository be located within their vicinity. We identify approximately one hundred of these discrete seismic reflection events on all six crooked seismic profiles. We interpret these events based on their seismic expression and apparent dip, enabling us to track them in 2D space. It is possible, and likely, for a single mafic dyke to exhibit a distinct seismic signature on multiple seismic lines. It is also possible that some of these reflections may correspond to unidentified intrusions or could be attributed to out-of-plane energy or coherent noise. Therefore, the presence of around one hundred reflections does not necessarily indicate the existence of one hundred dykes or even structures, as their 3D geometry remains unknown.

4.1.2 Vertical Seismic Profiles

The structural interpretation of the processed and migrated VSP profiles involves determining the three-dimensional positions and orientations of the imaged reflectors. This is accomplished by incorporating the coordinates of the shot points, borehole data, and the determined velocity within the processing methodology described in Sect. 3.1.5. Our interpretation strategy involves two approaches. Firstly, we utilize hyperbolic traveltime fitting to interpret coherent reflections on the processed depth-time VSP profiles (Cosma et al., 2003). Secondly, we rely on qualitative interpretation of reflections on VSP migrated sections. During hyperbolic reflection fitting, it is not possible to achieve complete 3D target localization from a single shot point due to the absence of the third dimension. To overcome this limitation, subsets of shot gathers are interpreted collectively using all the VSP processed sections from all three boreholes, where applicable. By cross-validating events observed in multiple profiles, the locations and orientations of the reflectors are determined (Cosma & Heikkinen, 1996).

The hyperbolae-fitting approach is demonstrated in Fig. 10a on a processed time-depth VSP for shot point 82 of IG_BH06. Note that not all the interpreted reflections result in the formation of a 3D reflector. As mentioned previously, only those reflections that correlate with other time-depth profiles can be defined in a 3D context. Similarly, Fig. 10b illustrates a migrated section in depth, where the vertical axis depicts a linear representation of the borehole trace along the receiver array. The horizontal axis represents the distance from the wellhead, with a south-southwest direction. Within this figure, several reflections associated with near-surface noise are denoted by solid green lines between the surface and a depth of approximately 400 m. Additionally, for reference, we have interpreted four deeper reflections that intersect with IG_BH06, which are illustrated as dashed red lines in Fig. 10b. These reflections align with the intersections of mafic dykes and the borehole trace, as documented during the drilling and coring operations of IG_BH06. There are a few other not interpreted but notable reflections located at distances greater than 50 m from the borehole. These additional sub-horizontal and shallow dip** reflections could potentially indicate the presence of further intrusions that were not sampled by the borehole. The interpretation of the migrated VSP sections is constrained by aperture limitations as observed by the lack of data beyond 500 m in Fig. 10b. However, these sections are utilized in conjunction with the reflectors interpreted on the processed time-depth profiles and surface seismic transects to produce a more complete structural interpretation of mafic dykes.

Fig. 10

IG_BH06 VSP interpretation through a reflection hyperbola-fitting on processed time-depth sections, and b south-southwest migrated section in depth. Line colors in a) and b are not related, green dashed lines represent fitting curves, solid green lines represent migrated reflections without geological significance, and red dashed lines represent migrated reflections associated with mafic dykes intersecting the borehole path

4.2 Seismic Response and Interpretation of Mafic Dykes

We use the available seismic information to establish the spatial character of the geological structures within the study area, with a specific emphasis on investigating the architecture of mafic dykes. By incorporating information from borehole data and surface geology, we have developed a comprehensive understanding of the properties of these intrusions, including their geochemical, geomechanical, and overall physical attributes, as discussed in Sect. 2.2. However, due to the limited number of available boreholes and their one-dimensional nature, the identification of the spatial distribution of mafic dykes has been challenging. To address this limitation, we aim to better understand their seismic signature by employing synthetic seismogram modeling that leverage the physical properties derived from well logging data. The 1D convolutional modelling approach used in this study focuses solely on primary reflections. The model assumes a time-invariant source pulse that propagates as a plane wave with normal incidence, and factors such as multiple reflections, head waves, and transmission losses are not considered. While acknowledging that full waveform modeling would offer a more comprehensive representation of the complete seismic response, our focus with this exercise is on identifying and isolating the signature of primary events. By analyzing the timing, amplitude, and waveform characteristics of the convolutional model, we aim to enhance our understand of the primary reflections associated with mafic dykes. This understanding aid us in interpreting their seismic expression in areas where borehole data are unavailable.

4.2.1 Seismic Signature of Mafic Dykes at the Revell Site

The mafic dykes exhibit a range of true thicknesses, varying from 0.05 to 3.5 m. Due to their small individual thicknesses, these intrusions are considered sub-seismic in nature. Consequently, they cannot be adequately imaged using conventional reflection seismic techniques unless specific criteria are met. These criteria include factors such as contrasting composition, significant impedance contrast, and the presence of vertical stacking or bulk structural arrays. These shallow dip** structures that are theoretically sub-seismic, meaning their vertical resolution is well below 1/4th of a dominant wavelength, may not always exhibit seismic transparency. Multiple studies have provided evidence that the imaging of small-scale features, even those with thicknesses as small as 1/30th of a wavelength, is feasible. Albeit not individually resolvable, these small-scale features may still be detectable (Brown, 2011; Dimmen et al., 2023; Hammer et al., 2004; Juhlin & Stephens, 2006; Sheriff, 1985; Zhou & Hatherly, 2014).

Our findings indicate that mafic dykes do exhibit a distinctive seismic signature that is recognizable within the 2D seismic profiles. These sub-horizontal to gently dip** structures show a highly distinctive composition as demonstrated in Sect. 2.2.1. They sharply contrast with the predominant biotite-granodiorite-tonalite to biotite-tonalite composition of the host rock. Additionally, most of these dykes exhibit high density and are associated with high P- and S-wave velocity compared to the surrounding rock, creating substantial impedance contrasts. Furthermore, the mafic dykes in the study area commonly occur in clusters as evidenced by observations from well logs. The stacked arrangement of these intrusions, combined with their high acoustic impedance, plays a crucial role in generating strong seismic reflectivity, as elucidated below. Collectively, these characteristics make the mafic intrusions favorable seismic targets, as their distinctive attributes facilitate their imaging. Nevertheless, it is worth noting that despite these distinguishing characteristics, there are certain limitations to the imaging and interpretation of mafic dykes. The effective imaging of these relatively small-scale structures can be influenced and potentially obscured by lateral heterogeneity in dip, the brittleness of the surrounding rocks, variations in dyke shape, the depth of the target, background noise, and other related considerations.

Figure 11 displays a segment of seismic transect 1.2, featuring the superimposition of the borehole IG_BH01 density and P-wave velocity log. The zoomed-in view encompasses an elevation range of approximately − 140 to − 450 m.a.s.l. Within this section, discrete mafic dykes, identified from borehole data, are denoted by blue dashed lines. An example is the I2 set, which consists of mafic intrusions stacked over a vertical span of approximately 10 m. Ludwig et al. (1970) and Salisbury et al. (1996) suggested that in most crystalline settings, a minimum normal-incidence reflection coefficient of 0.06 is necessary for an apparent reflection to be generated. In Fig. 11, the impedance associated with I2 is approximately $1.84 \times {10}^{7}\mathrm{ kg}/{\mathrm{m}}^{2}\mathrm{s}$, which corresponds to a reflection coefficient of 0.099 in relation to the surrounding biotite-granodiorite-tonalite rock, which has an approximate impedance of $1.50 \times {10}^{7} \mathrm{kg}/{\mathrm{m}}^{2}\mathrm{s}$. The strong reflection observed within the zoomed window in Fig. 11 correlates with the high P-wave velocity and density of the I2 mafic intrusions. Despite having an overall thickness that is only about one-sixth of the dominant wavelength, the stacked array of I2 dykes exhibits significant reflectivity. This reflectivity can be attributed to the combined effect of the intrusions’ high density, high velocity, and their stacked arrangement, indicating that surface seismic imaging can detect the acoustic contrast generated by these mafic dykes.

Fig. 11

Seismic Line 1.2 displaying the overlay of the IG_BH01 density (left) and P-wave velocity (right) log and a magnified view

As elucidated in Sect. 2.2.2, some mafic dykes exhibit reduced velocity while maintaining high-density. One such example is the I4 set, which can be observed at an elevation of approximately − 280 m.a.s.l in Fig. 11. In comparison to I2, the mafic intrusion displays a decrease of approximately 10% in P-wave velocity, while the variation in density remains within 2%. As a result, the normal-incidence reflection coefficient is approximately 0.04, evidenced by a very weak seismic reflection at the borehole location around the I4 marker, almost at background noise levels. This reduction in velocity can be attributed, at least in part, to variations in the composition and mineralogy of the intrusions. Mafic dykes showing this velocity reduction are mostly identified below elevations of − 100 m.a.s.l (approximately 550 m below ground). As presented in Fig. 2, these dykes have distinct geochemical characteristics with respect to the host rock. However, within the mafic dyke group (ellipse), there are local variations that could influence the specific characteristics of individual intrusions or sets of them. These variations can contribute to lower P-wave velocities observed in localized areas. Likewise, variations in mineralogy and texture within the mafic intrusions can contribute to velocity variations. For example, the presence of grain boundaries with different mineral compositions, fractures or veins can cause velocity heterogeneity around the dykes. This heterogeneity can result in localized areas of lower P-wave velocity. Figure 4 provides evidence of such velocity variations around the I6 marker at a depth of approximately 890 m. This specific set exhibits high density and a decrease in velocity, coinciding with an increased frequency of fractures. We believe that the structural alteration caused by fracturing leads to seismic transparency, and as a result, the I6 set is not associated with a discernible seismic reflection. The combination of a decrease in P-wave velocity and high density in mafic dykes presents a challenge for our interpretation, particularly when they lack a distinct stacking pattern. Their contrasting physical properties make them straddle between seismic targets and sub-seismic features.

While the study area does contain other subordinate intrusive rocks, such as tonalitic dykes, they are primarily transparent in seismic data. This is indicated by their overall normal-incidence reflection coefficient footprint, which is approximately 0.007 with respect to the host rock.

4.2.2 Synthetic Seismogram Modeling

We generate synthetic seismograms in boreholes IG_BH01, IG_BH04, and IG_BH06, which are situated close to seismic lines 1.2, 4, and 5. Figure 12 illustrates representative synthetic modelling results from borehole IG_BH01 for three scenarios: utilizing only density information, utilizing only P-wave velocity information, and utilizing both parameters (i.e., acoustic impedance). Additionally, the figure includes the markers of mafic dykes and a real seismic data trace. For visualization purposes, both synthetic and real traces are replicated seven times. The real seismic data trace is obtained from stacking over a 100-m interval, where the geometry of reflections is maintained in the vicinity of borehole IG_BH01. This stacked trace serves as a representative seismic signature for comparison with the synthetic seismogram. To facilitate a clearer visual comparison of relevant reflections, a bulk depth shift of 12 m has been applied to all synthetic seismograms. The need for this shift arises from the localized velocity discrepancies within the area, which can cause a misalignment between the synthetic seismograms and the actual reflection events recorded in the seismic data. By performing this adjustment, we enable a more accurate and meaningful comparison of reflection patterns in the seismic-to-well tie analysis.

Fig. 12

Synthetic well logs for borehole IG_BH01, and L2-normalized seismic trace. Synthetic logs are convolved with a 75-Hz Ricker wavelet for reflection coefficients derived for each the density (ρ), velocity (V_p), and acoustic impedance (AI) shown in Fig. 4. Mafic dyke occurrences are shown. Chosen frequency is comparable to the dominant frequency measured in seismic line 1.2

Among the seven intrusion intervals depicted in Fig. 12, a notable correlation between the synthetic model and the seismic trace is observed in five of them (I2, I3, I5, I7, and I8), i.e., over 70% of the mafic intrusion occurrences observed in this borehole are identified in seismic. The interpretation of the first interval, I1, has been excluded due to its unreliable nature. This interval is in the near-surface region, where imaging is affected by undesired arrivals. We interpret the seismic signature of I2 and I3 to be intrinsically related. Although both intervals seem distinguishable in all synthetic traces, they seem to appear as a single strong reflection package on the seismic trace and are trackable along most of the transect 1.2 (Figs. 9 and 11). This is to be expected as individual dykes may not be resolved but rather detected, strongly suggesting that similar mafic dykes across the study area are seismically trackable.

A particular set of dykes whose spatial distribution may be obscured during the interpretation of surface seismic data is I4. As described in Sect. 4.2.1, this set of intrusions shows a notable decrease in velocity while maintaining consistent density. This characteristic is evident in Fig. 12, where the velocity synthetic (V_p) shows limited sensitivity to I4, whereas the density synthetic (ρ) displays a pronounced contrast that generates a distinct reflection representing I4. In general, the seismic data provides limited detection of the presence of I4 in the immediate vicinity of the borehole. As we move away from the borehole and examine seismic transect 1.2, the weak seismic signature of I4 quickly diminishes to the level of background noise. Due to the combination of limited impedance contrast and the spatial orientation/geometry of these intrusions relative to the shooting line, they cannot be reliably traced using surface seismic alone. Nonetheless, our multi-source data integration approach, including VSPs and well log data, enable us to confidently identify the presence of similar dykes near borehole IG_BH04, IG_BH05, and IG_BH06, although with limited spatial resolution.

4.2.3 Interpretation of Mafic Dykes: Borehole Data, VSPs, and Surface Seismic

After identifying the seismic signature of mafic dykes using synthetic seismograms modeling, we conduct an integrated interpretation analysis combining VSPs, surface seismic data, and borehole information to unravel their structural nature and distribution. Within this section, we present a visual representation of the three-dimensional interpretative process, highlighting the correlation between logged mafic intrusions in boreholes and their corresponding reflections on surface seismic data and VSP migrated sections.

Figure 13a shows a north-east view of the seismic line 1.2, accompanied by a south-southwest migrated VSP section derived from borehole IG_BH05. The figure also includes color-coded markers representing different intrusion packages, labeled as intervals I1 to I8. To enhance the visualization of reflections, we show the pseudo-relief seismic attribute instead of raw amplitudes in Fig. 13 (Ercoli et al., 2020; Kumar & Sain, 2018). The markers visible on the surface seismic transect (Line 1.2), in Fig. 13a, correspond to borehole IG_BH01, located directly behind the transect in the depicted view. Figure 13b provides a rotated view of the same information. The remarkable correlation between dyke packages I2 and I3 as identified by the boreholes with the reflections observed on both the surface seismic and the VSP section is evident (indicated by the green arrows). The nearly orthogonal intersection of these seismic images, combined with the actual position of the mafic intrusions as identified in boreholes, provides us with insights into their three-dimensional character, including their dip and strike.

Fig. 13

a North-east view and b east-southeast view of pseudo-relief seismic transect 1.2, south-southwest migrated VSP section of IG_BH05, and colored markers indicating dyke occurrences along IG_BH01 (left) and IG_BH05 (right)—common dyke intervals to both boreholes share the same color; c southward view and d zoomed-in window of seismic transects 1.1 and 4 with the south-west VSP migrated section of Ig_BH04. The images demonstrate a strong correlation between the reflections observed in surface seismic data and migrated sections of the walkaround VSPs with the presence of mafic intrusions observed in boreholes IG_BH01, IG_BH04, and IG_BH05

The significance of a holistic analysis is shown by the I4 set. As detailed in Sects. 4.2.1 and 4.2.2, I4 does not exhibit distinct visibility in the surface seismic section as its reflectivity signature blends with background noise due to a decrease in seismic velocity around borehole IG_BH01. Nonetheless, the south-southwest migrated VSP section proves to be sensitive to this dyke interval, as indicated by the orange arrows in Fig. 13b. Not only does a continuous reflection (indicated by the orange arrows) exhibit a strong geometric and positional correlation with the I4 set of borehole IG_BH01, but it also matches with the mafic intrusions identified in borehole IG_BH05, as observed by the orange markers along the trace of both boreholes IG_BH01 and IG_BH05. Additionally, despite being weaker in comparison to the distinct reflection package associated with I2 and I3, this VSP reflection provides valuable insights into the lateral extent of the I4 set, which remains undetectable by surface seismic methods. It is noteworthy that the primary factor contributing to the imaging discrepancy between the VSP and surface seismic images is the significant difference in frequency content between the two datasets. Specifically, the VSP data exhibits a frequency content that is approximately three times higher in this depth interval compared to the surface seismic data. The VSP image shows a strong correlation with borehole markers among the other intrusion sets as well, except for I5. This intrusion set displays a locally steep and potentially oblique orientation with respect to the migrated VSP section. As a result, its imaging is hindered, as demonstrated by previous VSP numerical experiments conducted by Reiser et al. (2017) and others.

Another instance of multi-source data interpretation is observed in Fig. 13c and the zoomed-in window in Fig. 13d. In these figures, a distinct VSP reflection can be observed, aligning with shallower mafic intrusions identified along borehole IG_BH04, more clearly in Fig. 13d. This VSP reflection also exhibits a spatial correlation with a pronounced reflectivity pattern on surface seismic Line 1.1 and Line 4, as indicated by the dashed-red line. The observed fading of the VSP reflection away from the borehole, more apparent in Fig. 13d, can be attributed to limitations in migration caused by the restricted coverage of the VSP acquisition geometry.

The integrated workflow of generating synthetic seismograms, and incorporating additional data sources, namely VSPs and surface seismic, help us perform and validate the interpretation of mafic dykes beyond the borehole locations. By comparing the synthetic models with real seismic data from multiple sources, we successfully identify and correlate the seismic signatures of mafic intrusions in multiple scenarios. Despite challenges such as limited resolution and obscured spatial distribution of certain intrusions in surface seismic data, our integrative approach allows us to confidently identify seismic-scale mafic dykes throughout the field. The outcome of this workflow produces 3D surfaces representing the distribution of mafic dykes within the study area, as further detailed in the subsequent section.

4.3 Spatial Distribution of Mafic Dykes

The complete 3D character of a total of 31 mafic dyke sets has been successfully interpreted and correlated at the Revell Site. Each of these sets is defined by a distinct surface that may or may not correspond to a single mafic intrusion. However, it is probable that most of these surfaces depict the seismic response resulting from a stack of dykes. Figure 14 displays a northwest view of the Revell Site with the six pseudo-relief seismic sections and boreholes, along with the spatial distribution of mafic dykes. Note that due to the 3D perspective and the geometry of the seismic lines, not all surfaces representing mafic dykes are visible in the figure. Among the 31 dyke sets, 22 are identified using the VSP approach described earlier, while the remaining 9 are identified using surface seismic data. There is a fundamental correlation observed between the dyke sets identified on VSP data and those observed on surface seismic data. However, the latter, with its limited resolution and geometrical disposition, only allows for the retrieval of the 3D characteristics of approximately 30% of the identified mafic dykes.

Fig. 14

Interpreted surfaces derived from surface seismic data (green surfaces), and VSP reflection surfaces (blue surfaces). Surfaces are interpreted to represent individual or clusters of mafic dykes. Northwest view

These seismically derived mafic dyke sets demonstrate a strong correlation with the orientation and geometry of those dykes observed in the borehole data. Stereonets depicting the poles to mafic intrusions logged along the six boreholes (Fig. 3) and interpreted from seismic data (Fig. 15) are presented. The mean dyke dip observed in boreholes, documented as 24° towards the northwest (348°) in Sect. 2.2.1, aligns closely with the dip observed in the seismic datasets, which measures 12° towards the northwest by west (335°). The subtle disparity in dip and orientation is possibly due to the differences in the scale of observations. The borehole data captures a wider range of intrusions, including sub-seismic intrusions with varying geometries. The larger number of intrusions identified in the boreholes compared to the seismic data suggests that the sampling from the boreholes provides a more comprehensive representation of the intrusions in the subsurface, introducing a potential bias in the comparison. Note that in Fig. 3 we have opted against exclusively presenting seismic-scale intrusions identified in the boreholes, as such an approach could potentially introduce new biases. While the seismically interpreted dykes primarily exhibit a north-oriented trend, Fig. 15 illustrates the remarkable similarity between the range of geometrical variations observed in mafic dyke sets identified on VSP data and those identified on surface seismic. This is evident from the distribution of blue and green poles representing these sets. The variation in orientation is influenced by the predominantly gentle dip** nature of the intrusions. Some of these dykes exhibit nearly horizontal orientations, posing a challenge in accurately assigning their true attitude.

Fig. 15

Stereonet of poles to surfaces interpreted from 2D seismic reflection and VSP surveys. The orientation of interpreted surfaces is consistent with mafic dyke orientations in Fig. 3. Pole density contouring using a fisher distribution with a 5% count circle area

The imaging of mafic intrusions provides compelling evidence that they exist as discrete entities within the study area, as depicted in Fig. 14. While their spatial distribution is believed to extend across the entire region, the focus of imaging efforts has been primarily on areas where boreholes and seismic data are available. It is noteworthy that most identified intrusions are associated with a distinct rock domain, suggesting a potential genetic relationship between the intrusions and the surrounding rock. However, investigating the mechanisms underlying their emplacement falls outside the scope of this paper. Instead, in the following section, we delve into a discussion of the rock domains observed at the Revell Site and their relationship with the occurrence of mafic dykes.

4.4 Seismic-Constrained Interpretation of Domain Boundaries

Geological domains are distinct 3D regions within a larger geological system that share similar characteristics or properties. The process of delineating bedrock domains requires a thorough analysis of available geological data, including borehole logs, surface map** and geophysical data to identify patterns, trends, and relationships between different rock units. The initial modeling of rock unit domains across the Revell Site is solely based on the analysis of borehole core samples and geophysical log data obtained from six deep boreholes. Domain intervals are defined to include borehole intervals that contain clusters of shallowly dip** mafic dykes, as well as any adjacent zones of influence exhibiting altered or fractured bedrock. Figure 16 presents rock unit (RU) domains intervals obtained from boreholes, viewed approximately along a southwest to northeast transect. The mafic dyke domains, classified as RU3a and RU3b, are highlighted in green. RU3a is interpreted in four of the six boreholes, defined at the first occurrence of a mafic dyke along the borehole trace. The thickness of this shallower domain ranges from 10 m in IG_BH03 up to 54 m thick in IG_BH04, suggesting this domain gradually thickens and deepens in a northeast direction. The RU3a domain is absent from boreholes IG_BH02 and IG_BH06. RU3b domains are interpreted in all six boreholes, occurring through the middle and deeper parts of the boreholes. These domains typically comprise 5 to 15 dykes with borehole domain lengths ranging from approximately 150 to 450 m. The upper RU3b domain boundaries in all six boreholes are marked by either individual mafic dykes with approximate thickness of 1.5 m, or by clusters of dykes ranging in thickness from 3 to 6 m. Each cluster comprises between 2 and 3 mafic dykes. Similar to the shallow RU3a domain, the RU3b domain gradually thickens and deepens in a northeast direction. The lower domain boundaries, in IG_BH02 and IG_BH03, are interpreted within the length of the boreholes. However, the lower RU3b domain boundary in the remaining boreholes is unknown, located at some depth below the end of hole.

Fig. 16

Distribution of rock unit (RU) domains in six cored boreholes along an approximate southwest to northeast transect. RU3a and RU3b domains comprise clusters of mafic dykes with variable thickness

Results presented in Fig. 16 enable rock unit domains to be modeled spatially across the site. Initial domain modelling involved interpolating boundary markers between six boreholes across the site. The separations between borehole boundary markers typically vary from approximately 500–2200 m, and distance from markers to model domain boundaries can be up to approximately 3000 m. As a result, domain boundaries are interpolated over large distances without any additional subsurface constraint, producing rather suboptimal results in terms of accurately capturing the geomorphological variability of the domain surfaces. This is evidenced in Fig. 17b and c where the profile of the upper RU3b boundary is depicted in blue, representing a smoothed surface dip** approximately 5° towards the northeast. Because the lower domain boundaries extend beyond the end of the boreholes, the lower domain surface is poorly constrained. Given the limited data available, the lower domain boundary cannot be interpolated across the site with much confidence solely based on borehole data.

Fig. 17

a View of seismic reflection images and borehole traces through the Revell site. Interpolated upper and lower domain boundaries of the larger rock unit domain (RU3b) containing mafic dykes, as observed in drill core and seismic reflection. Top of model domain is 500 m above mean sea level. b and c Present results of interpolated domain boundaries for Lines 1.2 and 5 (respectively) using borehole domain interpretations only (blue trace) and using borehole domain and seismic reflection interpretations as subsurface constraints (red). The irregular nature of the integrated domain boundaries better follows the coherent seismic reflector geometry

The migrated seismic images shown above, and their derived attributes provide additional spatial subsurface constrains. These images correlate to either discrete or stacked occurrences of mafic dykes, confirmed through correlation of surface seismic with borehole synthetic seismograms and lithological logs. Figure 17a presents a southeast 3D view of the pseudo-relief seismic reflection data. Prior to integrating the seismic interpretation, the initial upper domain boundary of RU3b is in good agreement with a series of coherent seismic reflections that are observed across all six seismic profiles. It is apparent that the dyke clusters in proximity to the upper boundary of RU3b consistently show elevated acoustic impedance values relative to the remaining mafic dyke clusters. This relationship is demonstrated in Fig. 4 around the I2 set at approximately 665 m.b.g.s, and shown directly in Fig. 12 where acoustic impedance is used to derive the reflection coefficients for 1D forward modeling.

The interpretation of the upper RU3b boundary on the 2D seismic transects leads to the introduction of multiple nodal points into the model. These nodal points help further constraining the spatial modeling of the upper RU3b boundary, significantly enhancing its reliability across the entire site. Figure 17b and Fig. 17c display the improvements realized through data integration, where the initial domain boundary in blue is mainly shifted to shallower elevations along the northwest side of Line 1.2 and the northern side of Line 5. Likewise, seismic interpretation further enables modelling of the lower RU3b domain boundary across the site as a contiguous series of high-amplitude seismic events at depths beyond the ends of the boreholes (Fig. 17a—RU3b bottom). The variability of orientations of the interpreted mafic dykes is also highlighted, particularly in the north–south trending seismic Line 5. The upper RU3b domain boundary is nearly horizontal across the seismic profile and coherent reflections show an increase in apparent dip with depth for the lower RU3b boundary. As illustrated in Sect. 4.3, the deterministically-modelled mafic dykes are mostly present within the defined upper and lower limits of the RU3b domain boundaries. However, some portions of the interpreted mafic dyke surfaces extend beyond the confines of the domain.

5 Conclusions

This study successfully integrates multi-source seismic and borehole data to characterize mafic intrusions at the Revell Site in terms of their spatial distribution and geometry, as well as their associated rock domains. Our holistic approach provides confidence in the identification of these intrusions, demonstrating the significance of integrating both surface and vertical seismic profiles in identifying nearly sub-seismic features. While it is important to note that some mafic dykes may not be imaged by seismic, the comprehensive integration of data and analysis performed in this study allows for the successful identification of most seismic-scale mafic dykes within the survey footprints. The characterization of over 30 discrete sets of gently dip**, north-oriented mafic dykes highlights their significance in the context of characterizing a site and designing a deep geological repository. As illustrated in this paper, there is a correlation between the presence of mafic dykes and increased fracture intensity and alteration of the host rock. As such, understanding the locations and geometry of these intrusions are important considerations as their presence can influence the distribution of fractures, altered rock zones, and fluid dynamics, thereby potentially impacting transport properties. By incorporating the structural understanding derived from our study, we can develop meaningful representations of the fractured bedrock in order to support the development of realistic flow and transport models to assess the long-term safety and performance of a potential deep geological repository at the Revell Site.