Integrated geophysical imaging of rare earth element-bearing iron oxide-apatite deposits in the Eastern Adirondack Highlands, New York

Iron oxide-apatite (IOA) deposits, which belong to the “Kiruna-type” end-member of the class of iron oxide-copper-gold (IOCG) deposits, are of growing interest because they often host critical rare earth element (REE) mineral resources (Hitzman et al., 1992; Foose and McLelland, 1995; Long et al., 2010; Van Gosen et al., 2019). Their REE-bearing minerals often contain significant concentrations of heavy REEs (terbium to lutetium and yttrium), which are generally less common, more expensive, and critical for advanced technology applications. Economic extraction of these elements can be challenging, however, and IOA deposits are not currently mined for REEs (Long et al., 2010; Van Gosen et al., 2017). Nonetheless, as extraction technologies continue to be developed, so do studies of exploration approaches and ore deposit genesis models for IOA deposits.

Geophysical methods, particularly magnetic, gravity, and radiometric surveys, provide a way of delineating subsurface geologic features over broad areas and thus play a key role in the exploration and characterization of mineral deposits and their geologic framework. The magnetic field responds to rocks rich in magnetic minerals such as mafic igneous rocks, certain metamorphic rocks, and granites that contain magnetite. Gravity anomalies reflect density contrasts generated by differences in rock type such as mafic versus felsic rock. Radiometric data provide approximations for concentrations of K, U, and Th via gamma spectrometry (IAEA, 2003). These measurements represent sources within the upper 30–100 cm of the earth’s surface because gamma rays from sources deeper within the ground experience higher amounts of Compton scattering and thus energy attenuation before reaching the airborne sensors (Minty, 1997). These methods are thus very useful for surface geologic mapping in vegetated terrains. Rocks producing radiometric anomalies include certain types of gneiss, carbonate rock, those rich in potassium feldspars, or rocks that have undergone potassic alteration (Shives et al., 2000; Sandrin and Elming, 2006; Sandrin et al., 2007; Shah et al., 2017). We note that for Th and U, the radiometric measurements are typically described as “equivalent Th” and “equivalent U” (eTh and eU, respectively) because these elements involve multiple decay series.

Airborne magnetic and radiometric data combined with airborne or ground gravity data have been used to better image and understand the context of IOA and IOCG deposits with varying success, depending on the geologic setting and deposit properties. In the absence of sedimentary cover, radiometric data may highlight alteration. K anomalies associated with potassic alteration at IOA, porphyry, and volcanic-hosted massive sulfide deposits were observed in several Canadian locales, in Kiruna, Sweden, and in southeast Missouri, respectively (Shives et al., 2000; Sandrin and Elming, 2006; Sandrin et al., 2007, McCafferty et al., 2019). For other types of deposits, such as peralkaline intrusions and heavy mineral sands, REEs are often associated with Th-bearing minerals and correspond to radiometric eTh anomalies (Meleik et al., 1978; Force et al., 1982; McCafferty et al., 2014; Shah et al., 2017). Magnetic and gravity signatures of these deposits are more variable. In Kiruna, southeast Missouri, the Great Bear zone of northwestern Canada, and Candelaria (Chile), deposits were shown to be coincident with or near the edges of magnetic and sometimes gravity anomalies, attributed to high magnetic susceptibility magnetite and/or high-density magnetite and hematite within the deposits in the less magnetic and less dense silicic volcanic or sedimentary host rocks (Kisvarsanyi, 1981; Smith, 2002; Sandrin and Elming, 2006; Hayward et al., 2013; McCafferty et al., 2016; Ives and Mickus, 2018). In contrast, the Gawler Craton of Australia, where the world-class Olympic Dam and other deposits are located, has gravity highs, but magnetic anomalies are either broad and regional or they are offset from the deposit. In some areas, the accumulation of magnetite in shear zones and the destructive alteration of magnetite near deposits is thought to result in an absence of deposit-scale signatures, whereas in other areas the geophysical character of the anomalies suggests that they represent deeper intrusions, which may have provided the heat needed to drive fluid circulation when the deposits formed (Rutter and Esdale, 1985; Smith, 2002; Clark et al., 2003; Direen and Lyons, 2007; Austin et al., 2012; Funk, 2013).

The eastern Adirondack Highlands host numerous Grenville-age (approximately 1050–980 Ma), low-Ti, REE-bearing IOA deposits rich in magnetite (Figure ¹), providing an excellent opportunity to examine the geophysical expression of such deposits. Many of the Adirondack IOA deposits were mined for iron starting in the 1700s in “boom and bust” cycles until operations completely ceased in 1971 because iron mining activity within the U.S. became more economic farther west. There are thus are multiple records and data sets that can be used to ground-truth geophysical data (Kemp and Ruedemann, 1910; McKeown and Klemic, 1956; Buddington, 1966; Farrell, 1996; Lupulescu and Pyle, 2015). Accessory minerals containing REEs (mostly apatite but also monazite, xenotime, allanite, and other minerals) were typically left in situ or discarded in waste or tailings piles; thus, they are also a potential resource (McKeown and Klemic, 1956; Staatz et al., 1980; Valley et al., 2009, 2011; Long et al., 2010; Mariano and Mariano, 2012; Lupulescu et al., 2015, 2017; Taylor et al., 2019). Similar to other IOA deposits, heavy REE concentrations are elevated relative to light REEs, suggesting increased economic potential (Valley et al., 2011; Mariano and Mariano, 2012; Lupulescu et al., 2017; Taylor et al., 2017, 2019; Van Gosen et al., 2017).

In December 2015, we collected high-resolution airborne magnetic and radiometric survey data, ground gravity stations, ground magnetic susceptibility and gamma spectrometry measurements on outcrops, and density measurements, geochemistry and mineralogical analyses on hand samples (Shah, 2016, 2020; Taylor et al., 2018; Shah et al., 2019). We use these combined data sets to develop 3D views of individual IOA deposits, the underlying mineral architecture, and the regional geologic framework. The results are used to facilitate development of remote-sensing imaging and exploration tools and to gain a better understanding of variations in deposit distribution and their relation to the geologic framework.

The Adirondack Mountains in northern New York State form a broad, approximately 27,000 km² uplifted dome comprising mostly Proterozoic metasedimentary and metaigneous rocks of the Grenville province (Figure ¹). Most of the observed rocks were emplaced during a complex series of Mesoproterozoic orogenic events from approximately 1.3–0.98 Ga, including the Elzevirian, Shawinigan, and multiphased Grenville orogenies. These events resulted in (1) accretion of terranes to Laurentia, (2) emplacement of intrusive rock associated with arc magmatism, (3) emplacement of intrusive rock during periods of extension associated with orogen collapse, and (4) widespread metamorphism (McLelland et al., 1996, 2013; Rivers, 2008, 2012). The region was later covered unconformably by Cambrian to Ordovician sedimentary rocks that were subsequently mostly eroded away (Rodgers, 1971; Lowe et al., 2017). Apatite-fission track dating suggests that exhumation of the Proterozoic rocks began in the Late Jurassic to Cretaceous, possibly related to passage over the Great Meteor hotspot (Roden-Tice et al., 2000; Roden-Tice and Tice, 2005). Uplift continues in modern times, with limited geodetic work suggesting rates of more than 2 mm/yr (Isachsen, 1975, 1981). Tomographic studies suggest low seismic velocities beneath the Adirondacks, which in turn suggests asthenospheric buoyancy as a mechanism for the uplift (Yang and Gao, 2018). Today, the Adirondack Highlands reach elevations of up to 1629 m above sea level and are thus the highest elevation Grenville province rocks in North America.

The most abundant surface rock types are anorthosite, mangerite, charnockite, and granite (often referred to as AMCG) plutonic rocks (Figure ¹) that were emplaced during postorogenic collapse following the Shawinigan orogeny 1.18–1.15 Ga (Chiarenzelli and McLelland, 1991; McLelland et al., 2001, 2004, 2013; Hamilton et al., 2004; Rivers, 2008). Other rocks include older gabbro, granodiorite, granite, tonalite, and metasedimentary rock, and younger hornblende granite (referred to as the Hawkeye granite gneiss). Postdating all of these is magnetite-rich leucogranite, discussed in more detail below. These rocks experienced granulite-facies metamorphism approximately 1.15–1.05 Ga (Walton and De Waard, 1963; McLelland et al., 2001; Williams et al., 2019), but they are typically described in the literature by their igneous and sedimentary protoliths. They are bound to the east by Paleozoic sedimentary rocks along the Lake Champlain thrust fault, associated with the Taconic orogeny (Stanley and Roy, 1987; Hayman and Kidd, 2002). In various areas, these rocks are covered with Quaternary glacial deposits.

The leucogranite is the primary host for most of the low-Ti IOA deposits in the region (McKeown and Klemic, 1956; Foose and McLelland, 1995; McLelland et al., 2002; Valley et al., 2009, 2011) and is observed along the eastern, northern, and northwestern edges of the Adirondack dome (Figure ¹). Its magnetite content is consistently high except within a few centimeters of some IOA deposits (Postel, 1952; Hagner and Collins, 1967; Valley et al., 2011). Structural relations and geochronological studies show that the leucogranite is younger than most of the other Proterozoic rocks, although the date of emplacement is debated (Chiarenzelli and McLelland, 1991; Aleinikoff and Walsh, 2015; Chiarenzelli et al., 2017). Varying degrees of deformation are observed in the Eastern Adirondack study area, including a series of folds creating complex outcrop patterns of banded marble, paragneiss, and intrusive gneiss (Figure ²). This deformation includes major east-trending structures between Ticonderoga and Paradox and north–northwest-trending structures between Ticonderoga and Mineville that are parallel to the gneissic layering or foliation (Walton and De Waard, 1963). In some areas, folds are cross-cut by younger pegmatite bodies and dikes (Lupulescu et al., 2011).

Although most of the iron oxide deposits are hosted in the leucogranite, some are hosted in other rocks such as the gabbro-hosted Kent, Tunnel Mountain, and Noble deposits and the paragneiss-hosted Vineyard and Butler deposits (Kemp, 1898; Newland and Kemp, 1908; Postel, 1952; Taylor et al., 2019). Postel (1952) notes that some of the low-Ti magnetite deposits occur in skarns, quartz veins, and pegmatite. There are also magnetite deposits with elevated Ti hosted in anorthosite, such as those at Sanford Lake near Tahawus (Newland and Kemp, 1908; Hartnagel and Broughton, 1951; Woodruff et al., 2013), and in gneiss, such as the Benson Mines (Lupulescu et al., 2014).

The Adirondack IOA deposits are believed to have formed during postorogenic extensional collapse in the latter part of the Grenville orogeny (approximately 1050–980 Ma) (Valley et al., 2011; Chiarenzelli et al., 2017; Regan et al., 2019). Multiple hydrothermal alteration episodes in the Eastern Adirondacks have been observed, including potassic alteration that was overprinted by later sodic alteration; all IOA deposits with significant REE concentrations are associated with sodic alteration (McLelland et al., 2002; Valley et al., 2011; Taylor et al., 2017, 2019). Such alteration, along with voluminous igneous activity and high heat flow, are considered a key component of IOA deposits worldwide (Hitzman, 2000). Although analogous iron oxide deposits in other regions, such as those in the Gawler Craton in Australia and the St. Francois Mountains in Missouri, contain IOA and IOCG deposits that may be genetically linked (Hitzman, 2000; Day et al., 2016), the Adirondack deposits generally lack copper and gold in economic quantities, and in this aspect they more closely resemble deposits in Kiruna, Sweden (McLelland et al., 2002; Valley et al., 2011; Lupulescu et al., 2015; Taylor et al., 2019).

Within the Eastern Adirondack study area, iron mines and prospects are clustered in roughly four areas (Figure ²). The largest area extends from south of Moriah to south of Elizabethtown and contains dozens of mines, including those in the Mineville area. Farther south, the area near Paradox includes multiple deposits. Two less developed areas lie close to Lake Champlain, one between Ticonderoga and Crown Point and the other on Split Rock Mountain near Essex. Abandoned mines, pits, mine waste, and tailings are present throughout the region. The largest tailings piles reside near Moriah and Port Henry; some were analyzed in detail by Taylor et al. (2019).

Information regarding the subsurface extent of the mines is important for the guidance and evaluation of geophysical models, but most of the mine entrances and shafts have now been filled with water and are inaccessible. Published reports describing the extent of ore bodies date to the late 1800s and early 1900s and are focused on mines toward the center of the study area near Port Henry and Moriah. The Cheever Mine was the first of the larger mines to be developed, probably because of its relatively easy access from Lake Champlain (Figure ² and ²); Kemp and Ruedemann (1910) estimate its opening to be approximately 1785–1790. The Cheever Mine shows north- to north–northeast-trending, west- or southwest-plunging lenticular pod-like deposits that can be traced for more than 1.5 km along strike (Kemp, 1898; Kemp and Ruedemann, 1910; Stoltz, 1911; Chiarenzelli et al., 2018). The deposit and host-rock layers flatten to a broad syncline to the west until they terminate at a fault, yielding an approximately 500 m wide deposit (Newland and Kemp, 1908). Approximately 3 km west of the Cheever Mine is the Pilfershire Mine, which Newland and Kemp (1908) note is also oriented north to north–northeast with a west to southwest plunge.

The largest and most productive deposits are those of the Mineville area, first developed in the mid-1800s (Farrell, 1996; Figure ²). Numerous clusters of pits and shafts tapped several long north–northeast-oriented lenticular or tabular ore bodies that deepen to the south, often becoming thicker with depth (Kemp, 1898). The cluster east of Witherbee is referred to as the Mineville group; these mines, which include the Old Bed and Joker mines described by McKeown and Klemic (1956), drew upon several magnetite-apatite bodies that were “proven for 1800 ft (approximately 600 m) on the long axis” with their southern extent not known (Kemp, 1898). In the decade preceding the mine closure in 1971, the Old Bed and Harmony mines had been extended to a depth of more than 945 m below the surface and more than 1115 m downslope (Farrell, 1996). North of Witherbee lies a second cluster referred to as the Barton Hill group, for which Kemp (1898) and Kemp and Ruedemann (1910) describe long chutes of ore that are generally parallel to each other and to the Mineville group. These deposits extend for more than 1 km near the surface and at least an additional 600 m underground. Smaller clusters include the Burt Lot and Fisher Hill mines, which Kemp and Ruedemann (1910) speculate were a northern extension of the Barton Hill group orebodies and the Smith and O’Neil mines. These deposits are also oriented north–northeast and deepen to the southwest, but they are not as thick nor as extensive as the Mineville and Barton Hill groups.

REE-bearing apatite is present within several deposits, but there is much local variation (Figure ²). Areas with higher REE concentrations include the Cheever deposits, the Mineville group, the Skiff mountain deposits, the Dog Alley Mine, and some mines at Dannemora, which is northwest of the study area (Kemp, 1898; McKeown and Klemic, 1956; Valley et al., 2011; Lupulescu et al., 2015; Chiarenzelli et al., 2018; Taylor et al., 2019). Notably, the Barton Hill group, which is less than a kilometer from the Mineville group (Figure ²), contains little apatite (Kemp, 1898; McKeown and Klemic, 1956). Most of the Hammondville mines are also low in REE-bearing apatite, except the Dog Alley Mine; the nearby Skiff Mountain mines contain generally higher amounts of REE-bearing apatite (Valley et al., 2011; Taylor et al., 2019).

The data used include those from airborne magnetic and radiometric surveys, ground gravity stations, and petrophysical measurements on outcropping rock and hand samples (Figure ³ and ³). These data are available from Shah (2016, 2020) and Shah et al. (2019). In some cases, we compared these data sets to geochemical data analyzed by Taylor et al. (2018, 2019). Analyses of the geophysical data include derivative maps, filtering, 3D modeling of the airborne magnetic data, and combined 2D modeling of the gravity and magnetic fields.

Petrophysical measurements on outcrops and hand samples (the locations are shown in Figure ³) were obtained to provide ground truth constraints for interpretation of the geophysical data. Sites were chosen to obtain multiple measurements on a variety of rock types, although limited road access, heavy vegetation, and rugged topography reduced accessibility in places. Rock types at each measurement site were identified using standard field methods and visually identifying mineralogy with a hand lens. For magnetic susceptibility on outcrops, we used either a Terraplus KT-20 or ZH Instruments SM-30 field magnetic susceptibility meter, noting that the precision of these meters is several orders of magnitude finer than the typical local magnetic susceptibility variability for these igneous and metamorphic rocks. At each station, we made multiple measurements and retained the median value. Radiometric measurements on outcrops were obtained with a GF Instruments Gamma Surveyor field gamma spectrometry meter using 2 minute measurements at each site. Density measurements were conducted on hand samples in the laboratory using standard immersion methods. A total of 268 radiometric measurements, 171 magnetic susceptibility measurements, and 70 density measurements were made over a variety of different rock types; data are available from Shah et al. (2019).

Airborne magnetic and radiometric surveys were flown in December 2015 by Goldak Airborne Surveys with 250 m line spacing and a nominal flying height of 125 m along west-northwest–east-southeast traverses (Figure ³). We summarize the main components of the airborne data collection and reduction here; details are provided by Shah (2016). Magnetic data processing included corrections for sensor and platform position variations, system noise, and diurnal field variations. Corrections for the international geomagnetic reference field used the 2015 model. Radiometric data processing included similar corrections as well as adjustments for cosmic and aircraft background radiation, background radon, spectral stripping, dead time, and flying height attenuation. Several targeted signal-to-noise ratio (S/N) enhancement approaches such as spectral filtering via singular value decomposition were applied. Tie lines were used to assist in the leveling of the data sets.

The data were collected along a draped surface with a nominal height of 125 m, but the height of that surface above ground was adjusted to safely maneuver the fixed-wing aircraft over the rugged terrain (Figure ³). Radiometric data were adjusted for attenuation using standard corrections (IAEA, 2003), but the magnetic field data were provided at the nonconstant flying height. In general, magnetic field data that are draped over topography at a constant height are preferred for interpretation so that anomalies due to changes in the sensor height are not misinterpreted as having geologic causes. A constant drape height becomes even more important when generating derivatives, filtered anomalies, forward models, and inversions so that height-difference anomalies are not propagated during these calculations (Nabighian et al., 2005). We thus applied a combined upward and downward continuation of the magnetic anomaly data to a draped surface with a constant height of 150 m above ground (Paterson et al., 1990). For areas where the flying height is lower than 150 m, this involves upward continuation, whereas for heights greater than 150 m, this requires downward continuation; for the latter, a low-pass cosine filter was used to prevent the amplification of noise. Reduction to the pole of the resulting grid (inclination = 69.1° and declination = −14°) was performed to align the magnetic anomalies with their sources. This shows little effect because the local field inclination is relatively steep.

The observed magnetic and gravity fields represent the integrated effect of sources over all depths, with shallower sources producing higher amplitude, shorter-wavelength anomalies, and deeper sources contributing to the longer-wavelength field (Blakely, 1996; Nabighian et al., 2005). IOA deposits that are shallow enough to be of exploration interest are likely to produce magnetic anomalies that are short-wavelength because of the size of the deposits in the region (less than 2 km in horizontal directions) combined with their shallow depth. Filtering magnetic anomaly maps to highlight short-wavelength anomalies is thus likely to enhance anomalies due to IOA deposits. We applied the matched band-pass filter anomaly approach by Phillips (2001) to the reduced-to-pole (RTP), constant 150 m drape magnetic field grid. This method provides not only a short-wavelength filter for the total field, but also an estimate of an “equivalent” depth for the source material, calculated by considering the mathematical relations between the source depth and the width of the filter. We note that because the band-pass-matched filter represents a range of frequencies, this equivalent depth actually represents a depth range. Furthermore, potential fields are nonunique, and if a source is wide, shallow, and somewhat uniform, then it could be interpreted as being deeper than it is in reality. However, magnetic sources that are uniform over long distances are generally very rare, so this scenario is unlikely with magnetic data analyses (in contrast, this can be a more common problem when applying such filters to gravity data because densities tend to show less spatial variability).

Equivalent depths are chosen by seeking “natural” changes in the slope of the log of the frequency spectra so that it can be fit in a piecewise linear fashion. For the Adirondack study area, the algorithm suggested a separation for equivalent depths of 140, 370, 940, and 3600 m. Depths are relative to the drape height of 150 m, so the anomaly with an equivalent depth of 370 m is likely to be of greatest interest (the 140 m equivalent-depth anomaly will mostly represent survey noise). This anomaly corresponds to a depth range of roughly 0–800 m below the surface.

The sizes and shapes of magnetic sources representing IOA deposits are of key interest for exploration and better understanding the deposits. We thus inverted the constant-drape magnetic anomaly for a 3D magnetic model of magnetic susceptibility using the method of Phillips (2014). For this method, the susceptibility is determined by minimizing an objective function describing the difference between the observed constant-drape field and the magnetic field calculated from the 3D magnetic susceptibility model. The latter is calculated over the model volume using the magnetic susceptibility model and earth’s local magnetic field via a 3D Fourier transform that allows for the inclusion of topographic effects. The potential field solution has an inherent nonuniqueness, so a weighting scheme for the magnetic susceptibilities is used. For these models, we assumed cubic depth weighting, which mimics the attenuation of the magnetic field with the source depth. It is important to note that a different weighting scheme would result in a different 3D susceptibility model that fits the observed data equally well.

We constructed two 3D magnetic susceptibility models, one focused on resolving features at the deposit scale, and the other covering a larger area to image the leucogranite. The model areas are mostly contained within the 2015 survey areas, but to achieve a rectangular horizontal extent, they needed to include small areas covered only by previous airborne surveys. We thus merged the new survey data with older regional data (U.S. Geological Survey, 2013) by first upward or downward continuing all surveys to a constant drape of 150 m (as described above), gridding the results, and then stitching the grids together by allowing vertical translations or tilts relative to a regional grid (Ravat et al., 2009).

The deposit-scale model covered a 41 × 20.5 km wide region in the central part of the survey area that includes the Mineville, Cheever, Pilfershire, and other mines. The voxel size was 80 m in the horizontal directions and 15 m in the vertical direction, with a total vertical extent of 1.8 km (reaching approximately 1 km below sea level). The regional scale model covers a 100 × 100 km area with 6.5 km vertical extent; the voxel size was 200 m in the horizontal directions and 100 m in the vertical direction. The total volume of the models was intentionally limited so that the inversion code could be run on a personal computer within a 24–72 h time frame. For these models, it is not impossible for the vertical extent to be smaller than the actual size of the magnetic bodies (a larger volume could have been used, but that would require poorer vertical resolution). This can potentially result in higher magnetic susceptibilities for the deepest voxels so that the resulting magnetic field more closely resembles the observed field.

It is important to recognize that the inversions provide a model only for magnetic susceptibility, so the solution effectively incorporates effects of remanent magnetization. Previous studies of similar magnetite deposits showed that magnetic susceptibility generally tends to dominate over magnetic remanence (Clark et al., 2003; Austin et al., 2014), but the contribution of remanence to the observed magnetic anomaly can nonetheless result in slightly higher modeled susceptibilities, assuming that the orientation of the remanent magnetization vector is similar to the observed field. If the rocks have a strong component of remanent magnetism that is oblique to the local field, the model is likely to produce spurious results. Self-demagnetization of rocks with high (>1 SI) magnetic susceptibility will also contribute to the magnetic susceptibility model (Austin et al., 2014), but it will generally work in the opposite sense by effectively reducing the observed anomaly and thus also reducing the modeled susceptibility.

With these caveats in mind, the 3D magnetic susceptibility inversion should nonetheless provide at minimum a qualitative view of the magnetic sources at depth, even if they may bear slight differences from reality in a quantitative sense. Where possible, we use historical data that provide other constraints on the dimensions of magnetic sources at depth to check the results of the inversion.

Gravity station data (locations in Figure ³) include previously collected data from the Pan-American Center for Earth and Environmental Studies database (University of Texas at El Paso Regional Geospatial Service Center, 2004) and more than 180 new ground gravity stations (Shah, 2020). Some of the previous stations were collected before GPS data were widely available, which can generate errors when estimating elevations in the rugged areas. We thus reprocessed the previous data using elevations from lidar surveys flown in 2014 and 2015 by the state of New York to obtain updated free-air and Bouguer anomalies. Spot checks using differential GPS at some of the new stations indicated that the lidar data provide reasonable elevations for the gravity reduction as long as the original horizontal position is reasonably well known (for this reason, some stations collected near steep topographic slopes were omitted because slight changes in the horizontal position resulted in large changes in the elevation estimates).

The new gravity station data were collected using a Scintrex CG-5 by making three 2 minute measurements per site and taking a weighted average based on the measurement variation over the 2 minute interval (Shah, 2020). The data reduction involved standard corrections for daily instrument drift by comparing measurements at a fixed base station at the start and end of the survey each day and assuming a linear drift, a tidal correction, a latitude correction (using a handheld GPS), and elevation and terrain corrections (using the lidar data grids). A complete Bouguer anomaly was calculated assuming a crustal density of 2.67 g/cm³.

Ground gravity station data collection can be challenging in the Adirondack Highlands because of limited road access, thick vegetation, and very rugged terrain; some areas are nearly inaccessible. The station coverage is thus very irregular, with station separation of up to 6 km or more in places (Figure ³). Although derivative maps, filtering, and inversion for density are of interest, applying these methods to such an irregularly spaced data set would generate artifacts that can be directly tied to station density. For this reason, we have limited gravity analyses to 2D regional modeling in combination with the aeromagnetic data.

Measurements of the magnetic susceptibility, density, and radiometric concentrations of the outcrops (Figure ⁴) each exhibit dependencies on the rock type. The magnetic susceptibility is very high (approximately 0.5–2 SI), as expected, for the magnetite-apatite deposits and for some tailings. For several measurements on the deposits, the field instrument was saturated at 2 SI, so the actual susceptibility values are likely to be even higher. Apart from the magnetite-apatite deposits, the leucogranite consistently exhibits the highest susceptibilities, with most readings in a range of $30×10−3−60×10−3SI$⁠. Other rock types show lower and variable susceptibilities, with the most variable being the pegmatitic rocks. Field observations and measurements suggest that some paragneiss is sometimes associated with high susceptibility. Although this is typically unexpected for paragneiss, examination of thin sections for these rocks showed that they locally contain as much as 1%–2% magnetite. Anorthosite exhibits the lowest susceptibility.

Radiometric thorium (eTh) is generally low except for the leucogranite, IOA deposits, tailings, and some orthogneiss (Figure ⁴). The eTh values observed in the IOA deposits, leucogranite, and tailings are especially variable. The charnockite samples generally show low eTh, but these measurements are all from the northern part of the survey area and might not be representative of the region (discussed below). Radiometric eU is generally correlated with eTh, with pegmatites also showing greater variability (not shown). Radiometric K is high for rocks expected to contain potassium feldspars, such as granite, charnockite, paragneiss, and in some cases pegmatites. Within the leucogranite, the K concentrations are highly variable.

The densities of most measured rocks fall within typical crustal ranges of 2.6–2.8 $g/cm3$⁠. Gabbro samples were not available for laboratory measurement, but this rock typically has a higher density of 2.8–3.1 $g/cm3$ (Sharma, 1997). The IOA deposits and pegmatitic rocks are also denser than the other rocks.

The RTP magnetic anomaly (Figure ⁵) is dominated by a >1500 nT high comprising a wide, roughly $12×12km$ zone near Ticonderoga with multiple 7–10 km long east–southeast- and east–northeast-trending branches toward the Paradox area and an approximately 3 km wide, approximately 20 km long north–northwest-trending branch toward the Mineville area. There are numerous linear features within the magnetic high that were not visible in the older data. The east-trending branch is composed of a set of prominent east–northeast- to east–southeast-trending lineaments that correspond to topographic variations, and the north–northwest-trending anomaly is also notably linear.

The broad magnetic high corresponds mostly to the magnetite-bearing leucogranite, including areas where the leucogranite is buried beneath glacial cover. The wide high near Ticonderoga appears to continue east into Vermont, where the magnetic source is presumably buried beneath the Paleozoic sedimentary rock (Figure ⁵). Highs covering less area are also observed west of Mineville, most of which correspond to local occurrences of gabbro or leucogranite. In some areas, where the leucogranite is adjacent to paragneiss, there is a clear magnetic contrast with low magnetic field values for the latter, but in other areas that correlation is not apparent (Figure ⁵ and ⁵). This may reflect magnetite-rich paragneiss or the presence of magnetic rocks at shallow depths beneath the paragneiss. The anorthosite is expected to be mostly nonmagnetic, but some areas show local magnetic anomalies; field observations suggest that these may represent gabbro that is not delineated on existing geologic maps.

The band-pass-filtered anomaly for an equivalent depth of 370 m (corresponding to sources between the surface and approximately 800 m depth) exhibits numerous long linear anomalies oriented in various directions (Figure ⁵ and ⁵), many of which are coincident with structures visible as gneissic layering parallel to foliation mapped by Walton and De Waard (1963). These anomalies presumably represent the complex tectonic history with multiple episodes of faulting and folding that have juxtaposed rocks with different magnetic properties, locally creating linear magnetic contrasts. The most prominent anomalies are those within the leucogranite, probably because the rocks within that unit are more magnetized.

The band-pass-filtered anomaly also exhibits numerous local highs (Figure ⁵), most of which correspond to IOA deposits. We note that there are also local highs without corresponding known deposits, including at Hogan Hill, where lidar data suggest historical prospects, and the hamlet of Ironville, where no deposits are currently known (but may be present). One set of highs corresponds to a narrow body of leucogranite, indicating the need for caution and ground-truth information when interpreting such anomalies as prospective areas.

The orientation of the magnetic highs is variable; those near Mineville generally trend north–northeast (similar to observations of the subsurface extent of the deposits in that area), whereas those near Paradox are east-trending. Many of the band-pass-filtered anomaly highs are coincident with and parallel to longer magnetic lineaments, which, in turn, suggests that the deposits are oriented parallel to local structures (Figure ⁵); in numerous areas, these correspond to structures mapped by Walton and De Waard (1963).

Where the subsurface extent of the deposits is documented (Figure ⁵ inset), the magnetic highs are present over their length. For example, separate north–northeast-trending anomalies are observed for the Barton Hill group and the Mineville group, consistent with the orientation of deposits described by Kemp (1898). The 250 m flight line spacing is, however, too coarse to image the deposits in full detail.

The deposit-scale 3D magnetic susceptibility model (Figure ⁶) shows high values (100 − 300 × 10⁻³ SI) in areas where deposits are present, similar to the high-pass filtered anomaly maps. Some deposits are associated with a greater volume of high susceptibilities than others, such as those in the Mineville area (Figure ⁶). Other deposits such as the Cheever and Campbell deposits (Figure ⁶ and ⁶) are somewhat smaller in volume.

The 3D models suggest plunge directions for the deposits. These directions arise from the frequency content of the observed magnetic field. For example, if a magnetic anomaly has higher frequency variations near its northern part than its southern part, the associated source will appear shallower to the north and deeper to the south. The models suggest that many of these deposits, including the Mineville, Pilfershire, and Cheever mines, plunge to the south and to the west, consistent with published documentation where available. However, other deposits, such as the Lou Smith prospect and the Kent deposit, appear more vertical, suggesting variability not only in the horizontal orientation of the deposits, as shown on the band-pass-filtered anomaly map, but also in the vertical orientation.

The regional scale 3D model, with the color scheme adjusted to highlight magnetic susceptibility values greater than $30×10−3SI$ (based on the petrophysical measurements), shows the inferred shape of the leucogranite (Figure ⁷). The inversion portrays a large, thick body (or a combination of multiple smaller bodies) near Ticonderoga that is more than 4 km deep and shallower (⁠$<2km$ deep), thinner “belts” that connect Ticonderoga with Paradox to the west and with Mineville to the north–northwest.

The most prominent features of the radiometric anomalies (Figure ⁸) are five rounded eTh highs near Port Henry. Each of these highs is associated with piles of mine tailings or landfill areas of different sizes. Geologic features are unlikely to contribute because the geologic map in Figure ² shows that these areas mostly correspond to glacial cover, which is not expected to produce large eTh or eU anomalies. In some areas, the eTh highs show linear extensions that correspond to roads; we note that in winter some tailings are used by the local community as a gravel resource for road maintenance. A comparison to geochemical analyses of samples from some of these piles shows that the eTh highs occur over tailings piles that are rich in apatite and thus also Th and REEs; other tailings piles that lack apatite do not generate significant radiometric anomalies (Taylor et al., 2019).

Apart from the large tailings piles, the airborne radiometric data generally reflect the surface geology (Figure ⁸). Large areas with low K, eTh, and eU mostly correspond to anorthosite, gabbro, or bodies of water. Areas near Lake Champlain with high K, eTh, and eU mostly indicate Paleozoic sedimentary rocks, many of which are carbonates. The eU concentrations are well correlated with eTh but are also much noisier (a common occurrence with radiometric data due to low S/N ratios for the associated U channels), so we primarily consider K and eTh anomalies.

The charnockite generally shows elevated K (probably representing potassium feldspar) but variable eTh anomalies. In the southern part of the survey area and near Mineville (Figures ⁹ and ¹⁰), the charnockite shows higher eTh values, but elsewhere the eTh is lower. In some areas, long, east–northeast-oriented linear eTh highs are present, such as west of Pharaoh Lake (Figure ⁹ and ⁹). These highs correspond to a linear topographic valley, suggesting association with an east–northeast-trending fault.

The banded mix of leucogranite, paragneiss, and marble shows especially variable K, eU, and eTh. Similar to the charnockite, eTh is higher in the southern part of the survey area and near Mineville (Figures ⁸–¹⁰). K values are generally low except at a few isolated areas, including several crescent-shaped areas near Paradox along the southern parts of Skiff and Bear/Potter Mountains and around parts of the mountains in the Hammondville area (here, local variations were verified in the field by Suarez et al., 2018); none of these anomalies are coincident with deposits (Figure ⁹). The K highs are also observed over most of Miller Mountain and near Crown Point along the eastern side of Bulwagga Mountain (Figure ⁸). Thin sections of samples from Bulwagga Mountain showed potassium-rich granites and an absence of sodic alteration. The eTh and eU values are correspondingly elevated at most of those sites (Figures ⁸ and ⁹).

Given the strong correlations between geochemical measurements of Th and REE concentrations, one might anticipate that deposits with higher concentrations of REE-bearing apatite would be associated with elevated radiometric eTh anomalies. Using geochemical data reported by Taylor et al. (2018) and La as a proxy for REE-bearing apatite (Taylor et al., 2019), we find that this association occurs in some areas but with multiple exceptions. For example, most deposits on Skiff Mountain are coincident with an eTh high and show higher La concentrations, whereas most deposits in the Hammondville area are associated with lower eTh and sample La concentrations (Figure ⁹). Exceptions in the Hammondville area include a deposit northwest of the Dog Alley Mine, which has low La but is coincident with an eTh high, and the Dog Alley Mine itself, which has high La but is only proximal to elevated eTh values (we note that the other Hammondville deposits are also proximal to eTh highs). Additionally, the Harris Mine is rich in La but associated with relatively low airborne eTh values. Ground radiometric measurements show that the Harris Mine host rock has low eTh values except within approximately 50 m from the deposit, which is smaller than the resolution of the airborne surveys. Near Mineville, the airborne radiometric data reflect the glacial cover and tailing piles signatures (Figure ¹⁰) and are not helpful for establishing the presence or absence of such a correlation. The eTh highs observed near the Cheever, Pilfershire, and Mineville group mines are associated with especially large tailings piles.

In other regions, such as Kiruna and southeast Missouri, the ratio between radiometric eTh and K has been used to distinguish those K anomalies that are likely to be due to potassic alteration rather than primary igneous phases. In general, eTh/K is expected to be low in potassically altered areas because the alteration is likely to increase K relative to eTh (Shives et al., 2000; McCafferty et al., 2019). In the Eastern Adirondack study area, the eTh/K ratio is high near most deposits (Figures ⁸, ⁹, and ¹⁰). This holds true even in parts of the leucogranite where K is elevated, such as near the Dog Alley Mine and Skiff Mountain mines: although these deposits are proximal to elevated K, they are also coincident with eTh highs. Previous studies show that although potassic alteration occurred in parts of the region, most was overprinted by sodic alteration (Valley et al., 2011; Taylor et al., 2019). This sodic alteration may have decreased K relative to Th, resulting in higher values of eTh/K.

The complete Bouguer anomaly (Figure ¹¹) shows a broad, 8–10 km wide, approximately 50 mGal north–northwest-trending high that is aligned with, but much wider than, the outcropping set of leucogranite, marble, and paragneiss that encompasses the Mineville/Witherbee and Port Henry deposits. It also shows an approximately 5 km wide, approximately 40 mGal southwest-trending anomaly over the western part of the leucogranite, marble, and paragneiss set that encompasses the Paradox area. The latter anomaly might continue farther southwest over charnockite, but this is difficult to tell because the station coverage is much sparser in that remote area. The eastern part of the leucogranite, marble, and paragneiss outcrop set near Ticonderoga, which includes the Vineyard and other mines, is not associated with a gravity high.

The regional gravity anomaly shows limited correspondence with the surface geology and the magnetic field, with anomalies that are wider than the magnetic anomalies and the surface geologic units. Because the gravity field represents the integrated density variations over depth, it is more likely that the gravity variations represent deeper rocks. This possibility is further supported by the similarity in densities between most surface rocks (Figure ⁴). The only dense rocks are the pegmatites and IOA deposits, which are generally smaller than the station spacing, and the gabbro, which is only sparsely observed near the surface.

To test whether a deeper density source for the gravity anomaly is feasible, we generated 2D forward models of the gravity and magnetic field along one profile crossing the Mineville and Barton Hill group deposits (Figure ¹²) and along a second profile crossing the Paradox-area deposits (Figure ¹²). For these models, we used the results of the magnetic susceptibility inversion, which suggests that the leucogranite extends 1–2 km beneath the surface in the Mineville and Paradox areas (Figure ⁷). We find that for both profiles, the gravity anomaly can be modeled assuming a 7–12 km wide body residing at roughly 2–6 km depth with a density of $2.9g/cm3$ and surrounded by other rock with a density of $2.67g/cm3$⁠. The magnetic field can be roughly modeled assuming a $<2km$ wide body from the surface to approximately 2 km depth with a magnetic susceptibility of $50−60×10−3SI$⁠, corresponding to the leucogranite and similar to the 3D models. For the deeper, dense body, we tested a range of magnetic susceptibilities and found that values between $1×10−3$ and $8×10−3SI$ fit the observed magnetic anomaly equally well. This wide range is feasible probably because this deeper body has a longer wavelength signature and thus contributes little to the shorter wavelength anomalies.

The results of the 3D regional magnetic susceptibility inversion and 2D forward modeling of the gravity and magnetic field show the feasibility of layered geology that is consistent with the gravity and magnetic data. In the linear zone connecting Ticonderoga with Mineville and Paradox, this model has approximately 2 km wide, 1–3 km deep linear belts of rock that are strongly magnetic (susceptibility $50−60×10−3SI$⁠) but show little to no density contrast with the surrounding rock, consistent with the leucogranite that has been mapped at the surface. These belts are underlain by wider (10–15 km), deeper (more than 2–6 km depth), and dense (⁠$+0.23g/cm3$⁠) rock, with low to moderate magnetic susceptibility (⁠$1−8×10−3SI$⁠). The latter are consistent with mafic intrusions, discussed below.

Quantitative processing of the aeromagnetic data provides much insight into individual IOA deposits. Band-pass filtering can be used to locate most deposits, and 3D models of magnetic susceptibility reveal additional information such as the plunge directions and relative volume. For this study area, both approaches yield results that are consistent with previous documentation based on mining activity, where available (Kemp, 1898; Newland and Kemp, 1908; Kemp and Ruedemann, 1910). Our data set shows the importance of such data processing: at first glance, the broad high of the total magnetic field does not immediately suggest locations of IOA deposits because the anomaly is dominated by the signature of the magnetic leucogranite host rock, which is more voluminous than the deposits. However, the magnetic susceptibilities of the deposits are 30–40 times those of the leucogranite host rock, so they do contribute an anomaly that can be distinguished with appropriate filtering. The success of this processing illustrates the importance of data collection that is of high enough resolution for filtering applications. We note that the deposits themselves may have more complex shapes than can be imaged even by the 250 m line spacing survey, and that more detailed imaging might be possible with surveys that have a lower flying height and closer line spacing.

The modeled magnetic susceptibility values of the IOA deposits range from approximately 0.1 to 0.3 SI, which is about an order of magnitude lower than that measured on outcrops (1–2+ SI). However, the IOA deposits are typically layered within the leucogranite (Figure ²), so that in some areas only a fraction of the rock (approximately 5%–20%) comprises the magnetite deposit material. The bulk susceptibility of the deposit area could then be 0.1–0.3 SI, or possibly less. This, in turn, suggests that the remanence and demagnetization effects might counterbalance each other so that the susceptibility values in the inversion are probably within the range of those of the actual bulk susceptibilities, and that the dimensions suggested by the models are probably reasonable.

Although deposit samples show direct correlations among apatite, REEs, and Th concentrations (Taylor et al., 2019), the radiometric data do not show a clear relation between eTh and deposits with REEs except over the mine waste tailings piles. The limited correlation is probably because variations in Th concentration occur at finer scales than what the airborne data can resolve, such as between the adjacent Barton Hill and Mineville deposits, which have very different apatite and Th content, or the Harris Mine, where elevated Th is present in the host rock for only an approximately 50 m radius. Interestingly, most apatite-rich deposits are proximal to eTh anomalies even if they are not coincident with them.

Several lines of evidence suggest that observed eTh anomalies within the leucogranite and charnockite were generated by hydrothermal alteration. Some of the anomalies appear to be linear and associated with structures (Figure ⁹ and ⁹), which would occur if these structures served as fluid pathways. Additionally, broad, regional variations in eTh are present for the leucogranite and charnockite, which have different ages (Figure ⁸). This, in turn, suggests regional sodic alteration led to the deposition of Th over a wide area. Future work examining the minerology and geochemistry of samples from a broader region will help to better evaluate this possibility.

The Kiruna and Southeast Missouri IOA deposits exhibit elevated radiometric K and/or low eTh/K, interpreted as due to potassic alteration (Sandrin and Elming, 2006; Sandrin et al., 2007; McCafferty et al., 2019). The Eastern Adirondacks are notably different because neither widespread nor localized K anomalies are observed at deposits, and eTh/K is generally elevated in areas surrounding deposits (Figures ⁹, ⁹, ¹⁰, and ¹⁰). The absence of K anomalies near deposits is not surprising because most studies of the local deposits have shown that sodic alteration has overprinted earlier potassic alteration (Valley et al., 2011; Taylor et al., 2019). One sampled deposit (the Campbell deposit) does exhibit geochemical evidence of potassic alteration (Taylor et al., 2019), but it is not associated with radiometric K anomalies. It thus seems likely that the sodic alteration overprinting was very widespread, and remnant potassic alteration at sites such as the Campbell deposit is a very localized phenomenon. Some small areas do have elevated K and low eTh/K, such as Bulwagga Mountain (Figure ⁸), where thin section analyses are consistent with elevated potassium but no deposits are observed. Elevated K and low eTh/K is also observed and less than 1 km from (but not coincident with) the Dog Alley Mine (Figure ⁹).

The band-pass-filtered magnetic anomalies suggest that the deposits are structurally controlled. These maps show that they are aligned with linear magnetic anomalies that likely represent structural features (Figure ⁵), consistent with field observations that the IOA deposits are typically parallel to the foliation plane (Figure ²). This supports a model based on observations in other regions where IOA/IOCG deposits formed along existing structures that served as fluid pathways or conduits for hydrothermal fluid circulation (Hitzman, 2000; Barton, 2014). We note, however, that in some cases, the alignment between deposits and observed structures could have developed during later tectonic events.

Linear features observed in the magnetic total field and the location of deposits relative to the 3D model of the leucogranite strongly suggest that larger structural features served as the framework for the regional hydrothermal mineralizing system. The leucogranite has several long (⁠$>10km$⁠), quasilinear belts connecting the Ticonderoga and Paradox areas and the Ticonderoga and Mineville areas that are aligned with structures visible as gneissic layering or foliations mapped by Walton and De Waard (1963), shown in Figures ², ⁵, and ⁷. There may have also been additional regions or belts of leucogranite that are now obscured by fault displacement. For example, the Cheever and Split Rock deposits, which are adjacent to Lake Champlain, appear to be hosted in isolated bodies of leucogranite (Figures ⁵ and ⁶). However, later movement along faults bordering Lake Champlain may have truncated the leucogranite. The Cheever deposit has been observed to be fault-bounded to the west as well (Newland and Kemp, 1908).

Layered geology over several kilometers in depth is suggested by the limited correlation between the gravity anomalies and the surface geology or magnetic field plus the fact that most rocks exposed at the surface have similar densities. The magnetic field data are easily modeled with sources within the upper few kilometers and show good correspondence with the surface geology, so the gravity anomalies most likely represent deeper geologic sources (Figure ¹²). This necessitates a layered geologic model in which a wide (⁠$>7km$⁠) body of dense rock lies at approximately 2–6 km depth and beneath the shallower, narrower (approximately 2 km wide) quasilinear extensions of leucogranite that connect Ticonderoga with Mineville and Paradox (Figure ⁷). Forward modeling shows that this layered geology is quantitatively feasible. At Ticonderoga, where the gravity high is absent, such dense rock is not present, consistent with 3D model predictions that the less dense leucogranite reaches depths below 4 km (Figure ¹²).

These deeper rocks beneath the leucogranite belts, which are dense and may also have moderate magnetic susceptibility, likely represent mafic intrusions (Sharma, 1997; Smith, 2002). This layered model is reminiscent of those derived for the Gawler Craton, where some gravity highs are believed to represent deeper intrusions whose associated heat may have been a driver for the mineralizing system (Austin et al., 2012), and is consistent with observations at other IOA deposits requiring igneous activity and high heat flow (Hitzman, 2000). The Adirondack gravity data thus support the presence of intrusive rock that acted as a heat source, consistent with the IOA framework systems proposed elsewhere.

Within the Adirondack system, most IOA deposits are observed toward the distal parts of linear magnetic anomalies representing leucogranite and associated structures and near gravity highs that likely represent mafic intrusions (Figure ¹²). Deposits near Ticonderoga form a partial exception because gravity highs are not observed there. We speculate that the occurrence of deposits toward the distal ends of arterial circulation pathways of the hydrothermal system may be related to critical changes in structural, chemical, or thermal traps at these sites that decrease metal solubility and facilitate precipitation.

The geophysical data image various aspects of the IOA deposits and their geologic setting:

1. 1)

   The magnetic total field reflects highly magnetic host rock and major structures that likely played a key role in the hydrothermal system.

2. 2)

   Individual deposits can be imaged by applying band-pass filtering to the total magnetic field. More than 20 such deposits were imaged; these vary in strike but are typically aligned with surrounding structural features, suggesting that emplacement was structurally controlled.

3. 3)

   Some 3D magnetic susceptibility models highlight the subsurface extent of the deposits, providing information on the plunge directions and relative volume. The model susceptibilities are within the range of those expected from outcrops. This suggests that quantities that were not modeled, such as remanence and demagnetization, may contribute much less to the observed field than susceptibility and/or effectively cancel each other out.

4. 4)

   Radiometric data primarily reflect the surface geology and several large tailings piles that are rich in REE-bearing apatite. Although the geochemical data show a correspondence between REEs and Th within deposits, we do not see the corresponding radiometric eTh anomalies over apatite-rich deposits. This may be because deposition of apatite and Th enrichment varies significantly at very local scales.

5. 5)

   Although radiometric data at other IOA deposits exhibit zones with high K or low eTh/K, reflecting potassic alteration, such zones are not apparent in the eastern Adirondack Highlands. Instead, eTh/K is typically somewhat elevated near the IOA deposits. This is probably because the potassic alteration has been overprinted by sodic alteration.

6. 6)

   Gravity anomalies show little correspondence with surficial and shallow features. They likely represent deeper, dense intrusive bodies. Most of the IOA deposits occur over such anomalies. No information regarding the timing of these possible intrusions is currently available, but if they were emplaced and/or still cooling at the time of mineralization, they could have served as heat sources for the hydrothermal system.

7. 7)

   Combining the gravity results with 3D models of the leucogranite suggests key components of the mineralizing system architecture. It appears that most of the IOA deposits occur toward the distal ends of major structures above gravity anomaly highs. This may reflect the precipitation of magnetite and apatite due to important differences in critical structural, chemical, and thermal traps at the distal ends of these fluid conduits.

The airborne surveys were flown by Goldak Airborne Surveys; we especially thank B. Heath for the thoughtful processing of the radiometric data in this challenging, rugged terrain. We thank A. Klein and M. Goldman, who provided excellent and essential technical support in the field and in the laboratory. M. Lupulescu, C. Taylor, and J. Aleinikoff contributed stimulating discussions regarding the Adirondack deposits and local geology. Permission to access various deposits and tailings areas was graciously provided and/or facilitated by Lyme Adirondack Timberlands, J. and A. Reale (the Cheever Mine), J. Curran (the Pilfershire Mine), T. Scozzafazza (the Town of Moriah supervisor), and P. Tromblee (Solvay); we also thank M. Lupulescu, who helped with access arrangements. Special thanks go to P. Tromblee and B. LaMoria at the Port Henry Museum, who taught us much about the region’s history and assisted with locating relevant historical documents. We thank A. Hofstra, W. Day, and several anonymous reviewers for helpful suggestions that improved the manuscript. This effort was funded by the U.S. Geological Survey (USGS) Mineral Resources Program in collaboration with the USGS National Cooperative Geologic Mapping Program. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government.

Data associated with this research are available from five different websites: https://dx.doi.org/10.5066/F72R3PT0; https://doi.org/10.5066/P9NJ4F92; https://doi.org/10.5066/P9E6P7S2; https://doi.org/10.5066/P9EEXCKI; https://gis.ny.gov/elevation/metadata.htm.

Anjana K. Shah received an M.S. (1990) in mathematics from New York University and a Ph.D. (2001) in earth and environmental sciences from Columbia University. She joined the USGS as a research geophysicist in 2007. Prior to that, she conducted geophysical research for the Naval Research Laboratory and Dynamics Technology, Inc. (now Raytheon). She specializes in using magnetic, gravity, and radiometric methods to address geologic questions regarding critical mineral resources and intraplate earthquake hazards.

Ryan D. Taylor received a Ph.D. (2015) in geology from the Colorado School of Mines. He has worked as a research geologist for the USGS since 2008. His general research interests include studying the geochemistry and geochronology of ore deposits, particularly lode gold and porphyry deposits of the western cordillera of North America, and deciphering the processes that result in ore deposition within a regional framework.

Gregory J. Walsh received an M.S. (1989) in geology from the University of Vermont. He is a Geological Society of America Fellow who has been a research geologist with the USGS since 1992. He manages bedrock geologic mapping activities for the USGS in the Northeastern United States. He specializes in the mapping, structure, and tectonics of complexly deformed rocks, the integration of geologic data with hydrogeologic and mineral resource assessment studies, and the use of GIS as a mapping and analysis tool.

Jeffrey D. Phillips received a Ph.D. (1975) in geophysics from Stanford University. He first joined the USGS in 1973 and is now a scientist emeritus. He has conducted geophysical research for nearly 50 years on potential field theory and its application to geologic problems, including water, mineral, and energy resources.