Ambient Noise Tomography: A Sensitive, Rapid, Passive Seismic Technique for Mineral Exploration

For years now, the challenge has been to discover new mineral deposits in terranes that are buried beneath postmineralization overburden. Added to this technical challenge is a society-wide recognition of the need to accelerate discovery of mineral resources that can underpin the technologies required for the energy transition. These challenges of depth and speed in mineral exploration must be tackled via our current geophysical, geochemical, and geologic methods, each of which have different capabilities to “see through” the cover and are associated with a different rate for data collection, processing, and interpretation (Fig. 1).

Breakthroughs in the development of exploration methods have occurred in the past that have enabled step changes in the way exploration is conducted, leading to new discoveries. An early example was the advent of airborne magnetic imaging in the post-Second World War period that revolutionized the way geologists were able to trace rock units in areas lacking surface exposure (e.g., Grant, 1985; Nabighian et al., 2005). A similar breakthrough came with publication of a magnetotelluric (MT) survey across the Olympic Dam deposit that showed deep-seated features projecting toward the upper crust (e.g., Heinson et al., 2006, 2018), which added to the power of continental-scale MT surveys (Jones et al., 1988) for detecting large-scale structures that have accommodated the growth and reworking of continents and their mineral systems. The advent of trace-element geochemical analysis also spawned new approaches to “vectoring” for deposits, for example by determination of key elemental ratios to guide exploration within hydrothermal mineral systems (e.g., Wilkinson et al., 2015; Sader and Ryan, 2020).

What are the methods that will improve discovery rates this decade? Here we outline one such method that we believe will be a contributor to future discovery: ambient noise tomography (ANT). ANT is currently not widely considered part of the exploration toolkit (Arndt et al., 2017), despite the method’s capacity to image the subsurface in 3-D. Mineral exploration geologists are generally not accustomed to thinking about the subsurface in terms of seismic velocity, compared to other physical properties such as magnetism or density, which are embodied in potential field data. Nevertheless, seismic velocity depends on both density and elastic strength, which largely relates to the mineralogy and structure of rocks. As a result, seismic methods that measure velocity, such as ANT, are able to image important geologic aspects of the subsurface that can add to our geologic understanding and impact exploration decision-making (Malehmir et al., 2018). The ANT method has undergone significant advances in the past several years, not only in terms of instrumental sensitivity to ambient noise, but also in satellite-enabled rapid data delivery and subsurface velocity model construction (Hand, 2014; Olivier et al., 2022). ANT now has the capacity to offer a step-change improvement in the way the crust can be imaged, and can be applied with great value in the exploration for new mineral deposits.

The basis of all seismic methods is the detection of reflection and refraction of elastic energy moving within the earth. Active seismic methods artificially induce seismic waves by setting various types of explosive devices, vibrating the ground with specially designed trucks, or, for smaller-scale applications, activities like dropping weights or hitting a plate on the ground with a sledgehammer. Passive seismic methods, in contrast, “listen” to the seismic noise found within the Earth itself, whether produced by the huge release of energy during seismic rupture of fault systems or, at the other end of the scale, the energy generated by activity such as waves, wind, and indeed human activity. Both passive and active seismic methods rely on seismic sensors to detect the waves of energy. The correlation, timing, and frequency of these waves across an array of sensors provide a window into the velocity structure of the subsurface.

There are a number of passive seismic methods, including horizontal-to-vertical spectral ratio (HVSR), earthquake tomography, and ANT. The HVSR technique detects the horizontal and vertical components of seismic noise to determine a fundamental resonance frequency at a certain frequency band, a function of the average shear-wave velocity and subsurface layer thickness (e.g., Nakamura, 1989; Picozzi et al., 2005). HVSR is often used to investigate thickness of weathering profiles, or relatively shallow sediment such as soils or sands (Trichandi et al., 2024), but has been used for investigating thickness of sedimentary basins in general up to depths of 3 km (e.g., Jakarta basin; Cipta et al., 2018). Earthquake tomography including, for example, local earthquake tomography (LET) relies on modeling of earthquake hypocenters to understand the subsurface velocity structure (e.g., Kissling et al., 1994; Biryol et al., 2013). This method is often deployed to image crustal structure in seismically or magmatically active regions (Paulatto et al., 2022; Ranjan and Konstantinou, 2024).

ANT is a tomography technique that relies on a range of natural and anthropogenic noise sources and is very versatile in its application, with depth sensitivity that ranges from the shallow to very deep, depending on the frequency ranges that are sampled. Considerable method development has occurred over the past 20 years (e.g., Weaver, 2005; Curtis et al., 2006), building on earlier pioneering work (Claerbout, 1968). Applications of the method span the full range of scales, from continental-scale arrays imaging the deep structure of the continents (Bensen and Ritzwoller, 2008; Saygin and Kennett, 2010; Ward et al., 2013; Hoggard et al., 2020) to smaller-scale arrays focused on fault zone studies (Mordret et al., 2019), detecting underground cavities or voids (Cárdenas‐Soto et al., 2020), critical zone studies (e.g., Oakley et al., 2021), and the delineation of and exploration for mineral deposits (Hollis et al., 2018; Li et al., 2020; Chen et al., 2021; Ryberg et al., 2022; Jones et al., 2024).

Despite the fact that many academic and government organizations have utilized ANT at various scales to investigate everything from whole-Earth structure through to individual magma chambers, the uptake of ANT applied to mineral exploration has been surprisingly slow. Indeed, the relatively low uptake of hard-rock seismic methods also applies to active seismic methods, despite excellent examples where 2-D and 3-D active seismic programs have had major impacts on mineral systems models and exploration programs (e.g., Eaton et al., 2003; Wise et al., 2016; Heinson et al., 2018; Schijns et al., 2021). As a recent example, at both the Olympic Dam and Oak Dam iron oxide copper-gold deposits, sparse 3-D reflection seismic models have shown a consistent pattern in which the hematite-quartz zone in the core of these deposits can be resolved as a distinct P-wave velocity anomaly that has been used to better understand the geometry of these deposits and provide additional targets for exploration drilling (Schijns et al., 2023; Norman et al., 2024). It is clear there is a lot of value that can be obtained by applying seismic methods, active and passive, to hard-rock environments (Eaton et al., 2003; Duff et al., 2012). The 3-D, continuous-velocity model that seismic methods provide is a valuable data set to improve geologic understanding of an area of interest and is certain to add value to exploration workflows that rely on machine learning for data-driven prospectivity mapping. These are exciting developments, no doubt; however, in practice for most companies, seismic methods remain an adjunct to the core geophysical data sets of the mineral exploration industry: potential field and electrical data.

The ANT method utilizes cross-correlation of wavefields recorded at multiple seismic sensors computed to approximate Green’s functions (Shapiro et al., 2005; Bensen et al., 2007). In this way each receiver effectively becomes a virtual active source that tracks surface waves (Rayleigh waves) traveling between it and every other receiver in the array across a range of frequencies. Depending on the depth of interest, the frequency spectrum recorded can range from low (<0.1 Hz) to high (>10 Hz); however, in practice, different sensors are optimized for specific frequency ranges. The cross-correlation process identifies phase delays between each receiver pair, which is a property of the average velocity over the travel path that is modulated by the variable seismic velocity of the subsurface. With a 2-D array of receivers, the 3-D seismic velocities of the subsurface can be modeled. Cross-correlation across long time series and between multiple receiver pairs is computationally intensive, and advances in computer power in the past decade have greatly enabled this process.

Seismic surface waves are dispersive, meaning different frequencies travel at different speeds. In addition, different frequencies penetrate to different depths and thus sample geology of varying seismic velocity. As a result, Earth’s structure is encoded in distinct frequency bands of seismic noise. By analyzing the dispersive characteristics of that noise, we gain information about the shear-wave velocity structure at various depths beneath the array.

Travel times and dispersion information can be fed into tomographic inversion algorithms that recover a 3-D velocity model of the subsurface. As with all geophysical inversions, the solution is nonunique; many possible models are capable of fitting the observed data. To manage this challenge, inversion frameworks that provide some measure of the spread of possible solutions should be preferred. However, these tend to bring their own challenges; computational cost and the complexity of geologic interpretation increase.

ANT studies deliver data on the shear-wave seismic velocity field of the subsurface, which can be modeled in three dimensions. These models depict zones of higher or lower seismic wave speed. Like any geophysical method, it is important to relate these features to mineralogical, lithological, or structural characteristics when producing a geologic interpretation of the model. In general, seismic velocity increases with density and also with depth (Fig. 2). Typically, more mafic or higher-metamorphic-grade rocks have higher seismic wave speeds than more felsic or lower-metamorphic-grade rocks (Salisbury et al., 2003; Greenhalgh and Manukyan, 2013; Junno et al., 2020).

A second factor to consider is the scale of geologic features that can be resolved. The larger the volume of different geologic bodies, the more easily those bodies can be resolved. By reducing the spacing between seismic sensors, greater resolution of smaller features may be achieved. However, this may come at the expense of depth penetration. Depth penetration is a function of the wavelengths recorded by the survey; longer wavelengths penetrate deeper into the subsurface. If the total number of available sensors is constant, then smaller spacing between seismic sensors will reduce the overall aperture of the survey, which in turn limits the longest wavelengths the array can record.

While resolution of features is a concern for any geophysical method, when approaching ANT data, it is important to also consider sensitivity. An ANT model is more sensitive when more ray paths are detected between each stations across a range of frequencies. The more cross-correlations, the better variations in velocity of the subsurface can be constrained. It is critical to recognize that ANT images the bulk velocity of a rock volume rather than individual reflective boundaries, such as is the case for reflection seismic methods. A range of assumptions are necessary in both the formulation of tomographic models and in the interpretation of those models to derive information on other physical parameters and geologic features. Useful discussions on the derivation and limitations of seismic tomography are given by Foulger et al. (2013) and Paulatto et al. (2022), among others. As always, interpretation of geophysical data requires integrating as many data types as are available.

For the purposes of highlighting how mineral exploration can benefit from the application of ANT in general, we outline the recent developments in ANT technology by Fleet Space Technologies under the name ExoSphere (Fig. 3). Some of the key advances made under this banner, we believe, offer some learnings that may be able to be extrapolated to other geophysical methods more broadly, and doing so will see efficiencies that will benefit mineral exploration.

Exosphere ANT uses in-house built, real-time compact seismic nodes (Geodes), which have dedicated sensitivity to low-frequency signals (Olivier et al., 2022). Exosphere Geodes use a 2-Hz high-sensitivity geophone and low-noise digitizer, which results in an instrument noise floor (<–200 dBm at 2 Hz) that is more than 30 dB lower below 5 Hz than nodes that are commonly used in the industry. This sensitivity means the focus is on collecting signal originating in the upper crust, typically <10 km deep, and has the effect of reducing data gathering time.

Geodes are equipped with a real-time direct-to-satellite Internet of Things (DtS-IoT) modem and edge processing capabilities, enabling rapid data transmission for cloud-based further processing. Satellite connectivity is a true breakthrough for geophysical surveying. This is estimated to reduce the time needed for a field campaign by more than 50% (Jones et al., 2024), allowing for data processing and S-wave velocity model construction to be undertaken even while the Geodes remain in the field collecting data.

Fleet Space’s experience is that 3-D models can be routinely delivered in a matter of days following survey completion. We believe the connectivity through the DtS-IoT modem connected to a dedicated satellite constellation is the way forward for ANT and, where possible, potentially other geophysical surveying methods. The capacity to collect and deliver data at speed is a shift for exploration practice with significant real-world outcomes. Importantly, the near real-time monitoring of data as it is being acquired means that the survey can be reconfigured while field crews are still in the survey area to focus on interesting subsurface structure, or to respond in the case of instrument disturbance. In addition, noise tests can be carried out early in deployment to determine the local ambient noise conditions, and survey design can then be updated if required, thus ensuring the highest possible data quality. The use of the Exosphere Geodes therefore tackles both of the major challenges of contemporary mineral exploration: imaging the Earth in 3-D and collecting data at speed to inform and expedite exploration decision-making.

Any technology that can reduce the number of drill holes required to make a discovery of a mineral deposit is a major cost savings. In the absence of prior drilling information, estimating the depth of target horizon for a new drilling campaign is typically done on the basis of inversion of geophysical data such as magnetics and gravity to estimate the depth to the feature of interest (e.g., Thurston et al., 2002; Goussev and Peirce, 2010). One challenge for this type of modeling is that both magnetic and gravity methods lack depth sensitivity, leading to highly nonunique solutions. Estimates of depth from magnetic data may also be impacted by the influence of remnant magnetization, which if undetected can lead to an underestimate of the magnetic field strength.

Ambient noise seismic methods are able to differentiate cover sequences from crystalline basement since the former is typically seismically slower than the latter. An example comes from the Stuart Shelf, South Australia, where ANT has been used to image some 800 m of Neoproterozoic cover succession to identify the unconformity horizon in the vicinity of a basement-hosted IOCG deposit (Fig. 4; Olivier et al., 2024). The ANT velocity model reveals the pronounced velocity transition at the unconformity and also images velocity contrasts within the basin sediments themselves.

Robust depth information obtained from ANT can be used to constrain inferences derived from geophysical data sets that lack depth sensitivity. For example, the cover thickness derived from ANT can improve density estimates by constraining the interface between cover and basement in gravity inversions. In this way, the ANT enables discrimination of gravity highs that represent merely the presence of topography on the basement-cover interface and those that result from buried dense rocks, for example those rocks that may be prospective for hematite-rich IOCG systems (Healy et al., 2024).

Hydrothermal alteration of silicate rocks, common in large mineralized systems, may produce a significant enough seismic velocity contrast with less altered or unaltered country rocks to be able to be imaged using ANT. In addition, the presence of stocks and feeders to porphyry systems that intrude across preexisting layering in country rocks and are often themselves hydrothermally altered by magmatic fluids may produce velocity contrast and structural zones that can be imaged by ANT. In practice, it is usually the wider “footprint” of alteration systems that produce zones of contrasting rock properties that can be imaged by geophysics at the camp to deposit scale (Li et al., 2020; Chen et al., 2021; Comte et al., 2023).

A shear-wave velocity model of the Caosiyao giant porphyry molybdenum deposit, China, reveals the relatively slow velocities associated with the alteration system (Fig. 5; Chen et al., 2021). This deposit’s host rocks are garnet-bearing leucogranulite and plagioclase gneiss that are intruded by Mesozoic syeno- and monzogranite porphyry stocks (Wang et al., 2017). The ore zones are characterized by extensive structural overprint of the host rocks, with abundant cataclasite and phyllic alteration (Wang et al., 2017). The ANT shear-wave velocity model reveals that the most altered and mineralized regions correspond to velocity lows, as do several of the major fault systems in the area, while the granitic porphyry intrusives correspond to velocity highs (Fig. 5; Chen et al., 2021).

A second example of low-velocity zones in ANT data associated with mineralized rocks comes from the Hillside IOCG deposit, South Australia (Jones et al., 2024). Copper-gold mineralization at Hillside is associated with magnetite, garnet, and clinopyroxene skarn-type alteration (Ismail et al., 2014). Handheld velocity measurements indicate the altered rocks have velocities around 2,000 m/s, which contrast with velocities up to 2,800 m/s in unmineralized metaigneous rocks such as gabbro and syenite that dominate the footwall of the deposit. The ANT shear-wave velocity model reveals a major intermediate velocity feature that runs parallel to a demagnetized zone to the west of the deposit (<3,000 m/s; Fig. 6A-C). Mineralization is associated with magnetite alteration and shows a distinct magnetic anomaly to the east of the low-magnetic-intensity structure (Fig. 6C); mafic rocks within the host rock package appear to control the location of the main, north-south–trending gravity high (Fig. 6D). The copper grade shells correspond to a zone of depleted shear-wave velocity adjacent to the steep velocity high (Fig. 6E-F). When coupled with existing potential field data, the ANT models also serve to image the structural geometry of the deposit to greater depths, and to areas north of the deposit that are yet to be fully explored (Jones et al., 2024).

The reduction in rock strength caused by weathering in supergene systems can also be imaged using ANT. In the highly weathered and covered terrane of the eastern Gawler craton, South Australia, ANT has been used as a method for targeting potential extensions of supergene mineralization (Reid et al., 2024). Troughs of low velocity that correspond to zones of deeper weathering and associated supergene mineralization appear to be structurally controlled, being a similar orientation to faults in the underlying Proterozoic basement as imaged by aeromagnetic surveys. This is an example of the utility of the ANT method in delivering information on the geologic structure in 3-D, and these low-velocity features have provided a focus for exploration drilling.

Mineral exploration is a scale reduction exercise, starting at the terrane scale and becoming increasingly focused toward specific prospective domains and, indeed, specific structures or features at the camp and prospect scale (Hronsky and Groves, 2008; Arndt et al., 2017). In practice, once initial focus regions are chosen and individual properties staked, exploration targeting aims to rapidly identify and test those targets that have the best chance of containing economically significant mineralization (Lord et al., 2001). What are the main geologic and geophysical techniques that exploration companies typically use? How do these techniques fit together into an exploration program, and from our perspective, what is the role that ANT can play to reduce the risk of exploration spend?

Consider the stages of exploration for an as-yet undiscovered deposit within crystalline basement buried beneath several hundreds of meters of unmineralized cover. The initial area selection at the regional and district scale will likely have proceeded based on a prospectivity analysis conducted by a project generation team. The data sets that are generally relied on at this stage tend to emphasize the presence of, for example, mantle-scale structures (Begg et al., 2010), known regional mineralized corridors, and likely some form of numerical prospectivity modeling (Skirrow et al., 2019; Lawley et al., 2021; Farahbakhsh et al., 2023). At this stage, we envisage a powerful role for camp-scale ANT, or district-scale if possible. This would have the benefit of achieving an accurate depth-to-basement model to rapidly screen for regions beyond economic mining depths and to validate structural interpretations provided by potential field data.

The first-order pieces of information required at the tenement scale are likely to be surface geologic mapping and other geochemical programs to investigate prospectivity (e.g., Harris et al., 2024), along with magnetic and gravity data to determine structure and lithological associations (Fig. 7). From these data, maps of basement lithology and associated structural interpretations coupled with predictions of mineral systems that may be present in the area can be built. Follow-up ANT surveys over specific areas of interest will also assist with structural interpretations and the location of the next phase of geophysical data collection and inform drill targeting.

As was the case when transitioning from regional to camp scale, prior to drilling, a more focused, prospect-scale ANT survey could provide significant value. Geophysical inversion of potential field data sets can be enhanced by the accurate cover thickness estimation derived from ANT velocity models (Healy et al., 2024). Furthermore, as variations in mineralogy produce changes in seismic velocity, vertical and lateral velocity changes that are in some way “anomalous” from the regional background in the prospect-scale ANT may assist with mapping buried fossil hydrothermal fluid systems. In addition, in some instances, the geology of the cover material may mask other geophysical methods; for example saline groundwater may mask conductive zones at depth. Although seismic velocity will be modified by the presence of groundwaters if saturation is high enough to impact the bulk velocity characteristics, the method can also utilize lower frequencies that can provide velocity information at depth, below groundwater-hosting horizons.

Electrical geophysical methods can also be applied to those areas of interest. Time-domain electromagnetic (TEM) surveys of various types are useful at this stage to further identify the presence of conductive bodies or structures at depth. Magnetotelluric (MT) surveys are also useful for identifying both the lithospheric and crustal architecture and also shallower conductive zones that may be related to fossil hydrothermal systems and can image steep geologic structures with remarkable definition (Selway, 2014, 2018; Heinson et al., 2018). With these data, predictions can now be made with respect to the magnetic, density, velocity and electrical properties of the target areas. The decision might then be taken to drill the “best” targets to ground truth the geologic interpretation, ideally based around skillful placement of drill holes to reduce the uncertainty of geologic information (Caers et al., 2022). The pathway toward discovery is the integration of geologic, geophysical, and geochemical data with the courage and financial backing to drill.

Innovation is one of the key ways to improve mineral exploration success (Koch et al., 2015). With the dual challenge of depth of cover and the requirement to find more deposits within a shorter amount of time, there is a need to embrace new technologies that can unlock new insights into the subsurface. The application of seismic methods to hard-rock mineral exploration has been progressing steadily over the past decade, with more companies signaling their willingness to undertake seismic surveys to image the subsurface. The application of ambient seismic methods and, in particular, ANT in hard-rock exploration geology appears to be gaining momentum across a range of commodities and geologic settings. While ANT itself is not a new technique, recent methodological developments now make it practical and cost-effective for exploration. In particular, the breakthrough ability of a new generation of highly sensitive seismic sensors that can connect directly to satellites and enable near real-time data gathering is an exciting development that potentially foreshadows the future direction of geophysical data collection.

In an exploration campaign, not every drill hole intersects mineralization, let alone economic mineralization. Any technique that can reduce the expenditure on drilling and improve technical decision-making is a valuable addition to the exploration tool kit. While much work remains to be done to refine our understanding of what subsurface velocity contrasts mean in terms of geologic structure, the pace of development and application of the technique will no doubt see a raft of new studies over the next decade. Future developments in multiphysics inversions that are able to integrate properties that span the full range of petrophysics are the way forward to extract the most information from ANT and other geophysical data sets. In addition, imaging of the subsurface needs to move in the direction of inference-orientated imaging as opposed to discovery-mode imaging, in which quantitative predictions of rock properties can be made (Tsai, 2023). Greater integration of detailed 3-D velocity models obtained through low-impact, reliable ambient noise methods and other geophysical constraints into the mineral exploration workflow will serve to improve the rate of exploration success.

Colleagues at Fleet Space Technologies are thanked for their input into the ExoSphere ANT program. We also gratefully acknowledge the support of the many exploration companies that have utilized Exosphere. Rex Minerals is specifically thanked for their permission to publish results from the ANT program over the Hillside deposit. Ben Kay assisted with files to construct Figure 4. Professor Chen Guoxiong (China University of Geosciences, Wuhan) kindly supplied the shear-wave velocity model of the Caosiyao porphyry deposit. Keenan Jennings provided valuable feedback on an earlier version of this paper. John Paul O’Donnel and Heather Schijns are thanked for their expert, thorough reviews that have greatly improved this contribution.

Anthony Reid is a geologist with 20 years’ experience in regional geology and mineral deposit studies and a Ph.D. in the tectonic evolution of eastern Tibet from the University of Melbourne. Having worked for many years for the Geological Survey of South Australia, he has experience with applying different geologic and geophysical methods to understanding geologic processes. Most recently, Anthony has been working as a consultant geologist. Anthony has published widely on geologic problems and is an Honorary Visiting Research Fellow with the University of Adelaide. He is currently head of geology at Fleet Space Technologies, working with the geoscience team to provide geophysical models and geologic insights for clients across the exploration and mining industry.