Nondestructive determination of the physical properties of Antarctic meteorites: Importance for the meteorite—parent body connection

Photogrammetry is a low‐cost, nondestructive approach for producing 3‐D models of meteorites for the purpose of determining sample bulk density. Coupled with the use of a nondestructive magnetic susceptibility/electrical conductivity field probe, we present measurements for the interrogation of several physical properties, on a set of Antarctic meteorites. Photogrammetry is an effective technique over a range of sample sizes, with meteorite bulk density results that are closely comparable with literature values, determined using Archimedean glass bead or laser scanning techniques. The technique is completely noncontaminating and suitable for the analysis of rare or fragile samples, although there are limitations for analyzing reflective samples. It is also flexible, and, with variations in equipment setup, may be appropriate for samples of a wide range of sizes. X‐ray computed tomography analyses of the same meteorite samples yielded slightly different bulk density results, predominantly for samples below 10 g, although the reason for this is unclear. Such analyses are expensive and potentially damaging to certain features of the sample (e.g., organic compounds), but may be useful in expanding the measurements to accommodate an understanding of internal voids within the sample, lending itself to measurement of grain density. Measurements of bulk density are valuable for comparisons with estimates of the bulk densities of asteroids that are suggested as meteorite parent bodies.


INTRODUCTION
The physical properties of small solar system bodies, such as moons and asteroids, help us to understand how these bodies were formed and have evolved through time.We can study their physical characteristics through conducting geophysical measurements, such as asteroid bulk density and internal structure, from orbital and flyby missions (Consolmagno et al., 2008;Elkins-Tanton et al., 2020).We can also determine the physical properties of recovered meteorite samples sourced from such parent bodies to help interpret the composition and internal structure of a body that has been studied remotely (Consolmagno et al., 2008).Such laboratory investigations include determining meteorite density, porosity, magnetic susceptibility, and electrical conductivity.These properties of meteorite groups are diverse, reflecting the wide range of compositions, formation conditions, and secondary processes that they have undergone (Consolmagno et al., 2006;Krot et al., 2014;Ostrowski & Bryson, 2019;Rochette et al., 2003;Wilkison et al., 2003).Physical properties of meteorites are also used to inform future space mission planning.For example, the bulk density measured for the mesosiderite meteorite group has been proposed to be a good analog for asteroid 16 Psyche, a hypothesis that the upcoming NASA Psyche mission will test (Elkins-Tanton et al., 2020).

Density
Bulk density considers the ratio of mass to bulk sample volume, where the volume is encapsulated by the sample surface including voids and pore space (Consolmagno & Britt, 1998;Consolmagno et al., 2008;Macke, 2010;Ostrowski & Bryson, 2019).Grain volume is a direct measurement of the volume of the sample, excluding internal voids.Grain density is the mean density of the minerals within this volume, which can also be determined from sample modal mineralogy (Consolmagno et al., 2008).As the mineralogy of a rock sample is linked to its formation and subsequent processes of modification, rock density measurements can be used to understand what processes have occurred on the meteorite's parent body, such as differentiation, brecciation, metamorphism, and aqueous alteration (Consolmagno et al., 2008).High bulk density measurements result from high abundances of dense Fe,Ni metal and/or oxides and sulfides: For example, pallasites and iron meteorites have bulk densities ranging up to 8 g cm À3 (Table 1).In comparison, rocky achondrites have low bulk density, for example, 2.61-3.37 g cm À3 for the howardite-eucrite-diogenite (HED) group (Table 1).Thus, bulk density measurements can be used to infer the sample's bulk mineralogical makeup and can be used to aid in classifying the sample (e.g., Macke, 2010).A given bulk density may arise from different combinations of a range of minerals and sample porosities, as well as being highly dependent on the accuracy of the volume estimate, meaning that using density as a meteorite classification tool is not definitive.By contrast, as it is dependent only on sample mineralogy, grain density measurements provide a more indicative understanding of sample composition than bulk density.
Bulk density measurements are important in modeling the internal structure and mineralogy of asteroids, properties that help make connections between meteorite groups and asteroid classes (Akridge et al., 1998;McCausland et al., 2011;Wilkison & Robinson, 2000).Density is also a parameter in determining rock properties such as mechanical strength, which is key to understanding the physical evolution of asteroids (Scheeres & Sánchez, 2018).Porosity, which is intrinsically linked to density, is an important factor in understanding processes such as asteroid compaction, lithification, breakup, and reassembly (Consolmagno et al., 2008;Soini et al., 2020).Together, density and porosity are key in understanding impact shock and terrestrial weathering processes (Consolmagno & Britt, 1998;Ostrowski & Bryson, 2019).These properties are also fundamental to studies of impactor survivability during impacts between parent bodies (e.g., asteroid to asteroid), as well as meteorite survivability during Earth atmospheric entry (Avdellidou et al., 2016(Avdellidou et al., , 2017;;Flynn, 2014;Popova et al., 2011).

Magnetic Susceptibility and Electrical Conductivity
Magnetic susceptibility is the measure of the extent to which a material will be magnetized when a magnetic field is applied (Consolmagno et al., 2008).It is a useful metric TABLE 1. Summary of literature data for physical properties of selected meteorite types.

Meteorite classification
Bulk density a (g cm À3 ) Grain density a (g cm À3 ) Magnetic susceptibility a (logχ) Vol% Fe,Ni metal b for measuring the abundance of magnetic minerals in a rock sample, which for meteorites typically includes Fe,Ni metal, Fe-sulfides such as pyrrhotite [Fe 1Àx S], and Feoxides such as w üstite [FeO] or magnetite [Fe 3 O 4 ] (Consolmagno et al., 2008;Folco et al., 2006;Gattacceca et al., 2004).The ranges for magnetic susceptibility of different meteorite classifications (Table 1) reflect the amount and nature of magnetic minerals present in these meteorites.The large ranges of magnetic susceptibility values within a meteorite classification reflect heterogeneous concentrations of magnetic minerals between samples.For example, magnetic susceptibility values of Logχ for LL ordinary chondrites in Macke (2010) range from 2.86 to 4.73 (where χ has units of 10 À9 m 3 kg À1 ).The overlap between these ranges for different meteorite classes and groups (e.g., between ordinary chondrites and pallasites) limits the extent to which we can interpret the magnetic mineralogy of these samples from magnetic susceptibility measurements alone.Measuring electrical conductivity of meteorites in conjunction with magnetic susceptibility allows us to distinguish between a sample that contains nonconductive, magnetic minerals and those that contain conductive metallic phase(s) (Rochette et al., 2003(Rochette et al., , 2008(Rochette et al., , 2009;;Uehara & Gattacceca, 2023).The variation in the proportion and connectivity of Fe,Ni metal, magnetizable oxides, and sulfides among meteorites of different groups is linked to the processes by which these materials originally formed and were subsequently modified.Thus, knowing the proportion of these phases within meteorite samples is a useful tool to understand their mineralogy, as well as aiding in meteorite classification.

Methodologies for Determining Physical Properties
Due to their rarity, and the ease with which samples can be contaminated in the laboratory environment, it is important that the pristine condition of collected meteorite samples is preserved as much as possible during laboratory curation and classification.Some methods for determining density and porosity, such as the glass bead method, can cause fragmentation and may cause sample contamination (Consolmagno et al., 2008;McCausland et al., 2011).Although measurements can be performed on smaller sample sub-splits, these may not be representative of the whole sample (Consolmagno et al., 2008).
In this paper, we investigate the physical properties of a suite of newly recovered Antarctic hand specimen-sized meteorites (Joy et al., 2019;MacArthur et al., 2022).We have developed a scalable analytical laboratory process for determining whole meteorite bulk density, magnetic susceptibility, and electrical conductivity that is nondestructive and noncontaminating.We use professional photogrammetry software to produce 3-D models of meteorites, from which we derive volume measurements and computed sample bulk densities.Photogrammetry uses two-dimensional (2-D) images to determine accurate information about the surface of an object.Our method was conducted using a commercially available camera to collect these images.We compare volume and density values determined from photogrammetry with those determined by x-ray computed tomography (XCT) and meteorite data from literature, to assess the validity of the photogrammetry method.XCT is a predominantly nondestructive method for investigating the internal structure and composition of rock samples by obtaining large numbers of parallel 2-D xray images through a sample, which are used to reconstruct a 3-D model of that sample (Krzesinska, 2011;Liu et al., 2017;Mees et al., 2003).The greyscale of each 2-D image used to reconstruct a 3-D model of the sample reflects an attenuation of x-rays passing through the meteorite which depends primarily on the density and mean atomic number (Z) of constituent materials.
Simultaneous measurement of magnetic susceptibility and electrical conductivity of hand specimens of meteorites can be made using a nondestructive, noncontaminating device called a MetMet (Gattacceca et al., 2004;Uehara & Gattacceca, 2023).We use a version of the device intended for use in cold weather called an AMetMet which has been built by the same team.The combined approach allows us to interpret physical properties in conjunction with knowledge of the internal structure of the meteorites from computed tomography 3-D reconstructions.
Bulk density measurements, combined with AMetMet measurements of magnetic susceptibility and electrical conductivity, aid classification of the samples.Our approach is relatively low-cost, noninvasive, and avoids large amounts of handling, helping to avoid contamination of meteorite surfaces and internal structures.Future work determining porosity from the XCT data collected might also allow us to determine grain densities, but this is not explored here.We use the data to compare sample properties with properties of potential parent bodies, and to discuss potential sources for the collected meteorites.

Weathering of Meteorites
Meteorites undergo weathering on Earth.This varies according to their terrestrial age, location, and environment.The most commonly affected minerals are Fe-Ni metals, which are the main magnetic minerals; thus, magnetic properties are a proxy for weathering state (Bland et al., 1996;Gattacceca et al., 2011;Steer et al., 2015).The alteration products formed, such as Fe-oxyhydroxides (including goethite), can also have magnetic properties of their own (Steer et al., 2015).
Grain densities and porosities of both Antarctic and non-Antarctic finds were significantly altered by terrestrial weathering, which reduces the porosities of the meteorites by filling with weathering products (Li et al., 2019).There is no significant change to the mass or outside shape, so there is no significant difference between the bulk densities of falls and finds (Consolmagno et al., 2008).Rochette et al. (2003) found that magnetic susceptibility of meteorite finds was systematically lower for each class of ordinary chondrites, corresponding to weathering alteration.
Ordinary chondrites recovered from a hot desert, that had experienced more weathering (grade W3), were found to have a chemical remanent magnetization, acquired on Earth, most likely on the formation of maghemite and magnetite as weathering products (Uehara et al., 2012).In contrast, those with little or no weathering (grade W0 and W1) had weaker natural remanent magnetization, consistent with the magnetization of meteorite falls thus preserving their original extraterrestrial magnetization (Uehara et al., 2012).

Test Samples
We tested our photogrammetry data collection and processing workflow on a range of samples of different size, mass, color, and shape.Test samples included a shatter cone (impactite) from the Ries impact crater, a 3 mm thick slice of L5 ordinary chondrite Northwest Africa (NWA) 869, and a fragment of the H5 ordinary chondrite Gao Guenie (Figure 1).We also performed the photogrammetry workflow on two wooden cuboids to help determine the precision of the volume measurements we make with this method (see "AMetMet Measurements" Section).
Testing of the AMetMet device was undertaken using a known suite of samples from several meteorite groups (Table 2).Test samples (Figure 2) include IAB iron Campo del Cielo, H5 ordinary chondrite Gao Guenie, L3-6 brecciated ordinary chondrite NWA 869, and LL5 ordinary chondrite Chelyabinsk (a single stone with complete fusion crust which obscures the proportion of ordinary chondrite and impact melt lithologies).Several pieces of each meteorite except Chelyabinsk, with a range of masses, were measured.We weighed the test samples using a Denver Instrument MXX-123 (precise to 0.001 g) to determine the sample mass.We also measured several terrestrial rock samples from Antarctica, provided by the British Antarctic Survey, to help understand the range of Antarctic lithologies that were encountered by the field team (see Joy et al., 2019).

Antarctic Meteorites
The Lost Meteorites of Antarctica project conducted its first field campaign to search for meteorites during part of the Antarctic summer in 2018-2019 (Joy et al., 2019).A second season was conducted in 2019-2020, consisting of surface meteorite collection in addition to subsurface meteorite searching using skidoo-drawn metal detecting panels (Joy et al., 2019;Marsh et al., 2020;Wilson et al., 2020).From the first season, 36 postulated meteorite samples were returned to the United Kingdom and transported to the University of Manchester for curation and formal classification (MacArthur et al., 2022).Eighteen meteorites from this first season were selected for our physical properties study, including a range of sizes, colors, morphologies, and AMetMet data collected in the field.Meteorites were recovered from both the Outer Recovery Icefields (OUT) and the Hutchison Icefields (HUT).The sample suite includes 13 ordinary chondrites (1 of which is LL, 6 are L, and 6 are H), two mesosiderites, one CM-an carbonaceous chondrite, one aubrite, and one eucrite (Table 4; Meteoritical Bulletin, accessed May 2022).

Laboratory Setup and Mass Measurement
Upon their return to the United Kingdom, the recovered Antarctic meteorites were initially stored in a freezer in the Isotope Group Class 1000 clean room laboratories at the University of Manchester.They were defrosted under low vacuum conditions in the clean laboratories to prevent water-rock interaction during the thawing process.Sample masses were measured using a KERN PLS high precision (precise to 0.001 g) scale.For samples with mass >1 kg, masses were measured using a KERN ECB scale (precise to 5 g).

Photogrammetry
Photogrammetry data collection was conducted in the clean laboratories at the University of Manchester.Meteorites were imaged on a plastic turntable, demarcated with 5°rotational intervals, in a white light tent, with LumeCube remote controlled diffused lights used to create a controlled light environment (Figure 3).The samples were protected from contamination by placement on a portion of the same sterile low-density polyethylene (LDPE) plastic bag that was used to store the samples.The samples were handled using type 304 stainless steel tongs to prevent potential metal contamination caused by the comparatively elevated highly siderophile element abundances in stainless steel tools, relative to naturally formed metals in pristine extraterrestrial samples such as lunar meteorites (Day et al., 2018).Tongs were cleaned with isopropyl alcohol (IPA) between uses to ensure that there was no cross contamination of material between the handling of different meteorites.
The sample photographs were acquired using a nextgeneration 45.7-megapixel (MP) Nikon D850 digital single-lens reflex camera (DSLR) with a 50 mm focal length lens.When using the largest file format, each image contains 45.7 million pixels available for software processing to identify as potential matches within other images, although we also produced successful models with other lower resolution cameras (e.g., 25-MP Nikon D3300).Camera settings were adjusted to very low  aperture width to maximize depth of field, and low ISO to maximize the amount of detail in every picture.White balance was also maintained for each suite of photographs to maintain consistency in the final product.Photographs were captured in manual mode to a RAW file format so that all the information is preserved, and any color balancing is applied equally (see Supporting Information to access these images).
Photographs were taken at 5°intervals around the sample, with the sample in two or more orientations (achieved by turning the sample over on the turntable) to ensure adequate coverage of the sample surface.Where there is concavity in the surface such as voids opened by cracked fusion crusts, photographing from multiple orientations improves the quality of the internal topography that the photogrammetry software can reproduce.In addition to the photography, the meteorite samples were measured along a principal axis, using vernier calipers, to calibrate the image scale once the model has been produced.
We performed preliminary processing of the images in Adobe Lightroom (V5.1), manually adjusting contrast, exposure, and white balance values of the entire image suite at the same time, to produce an accurately colored 3-D model.The images were then loaded into Agisoft Metashape Standard (V1.6.1)workspace.The first stage of processing, Align Photos, performed a preliminary pixel matching analysis to generate a sparse threedimensional point cloud (composed of e.g., $80,000 points), and to compute the digital camera positions relative to the sample (Figure 4a).This was then converted to a dense point cloud using the Build Dense Cloud operation (Figure 4b).The dense point cloud (e.g., >1,000,000 points) gives a good impression of how well the software has reconstructed the sample surface.After the dense point cloud was generated, we used the Build Mesh operation to build the shape of the 3-D model (Figure 4c).The mesh is a 3-D object file containing a record of the sample shape.The Build Texture operation was then used to produce a mosaic of images derived from all the input photographs that gives the model its real-life appearance (Figure 4d).Once these steps were completed, the shape file (e.g., .objor .fbxfiles) and texture file (e.g., .tiffor .jpegfiles) were exported and saved.For a more detailed explanation of 3-D model production using Agisoft Metashape, see Appendix C (Figures C1-C26).
The photogrammetry shape files were exported as .objfiles and imported into Autodesk 3DS Max, a professional computer graphics program.The 3-D sample models imported into 3DS Max, and the principal axes measured in the laboratory were manually aligned to the x, y, and z axes in the software workspace by rotating the model.The models were scaled to the correct dimensions using the 3DS Max scale with fixed dimensions to ensure that the aspect ratio of the model was maintained.The 3DS Max measurement tool was then used to compute a sample volume.To understand the uncertainty in measurements derived by photogrammetry, we produced 3-D models of two wooden cuboids of known size (5.10 × 5.10 × 10.10 and 2.50 × 2.50 × 10.10 cm, measured with the Vernier calipers).The smaller wooden cuboid had a known volume of 63.13 cm 3 and a computed volume of 61.87 cm 3 , which is a 1.99% relative underestimate.The larger wooden cuboid had a known volume of 260.1 cm 3 and a computed volume of 258.91 cm 3 , which is a 1.44% relative underestimate.Based on these measurements, we give an approximate uncertainty on photogrammetryderived bulk density values calculated from a 2% relative error on the volume.We propagated the errors for photogrammetry-derived volume and sample mass to determine an error for photogrammetry-derived bulk density using a standard approach described in "Bulk Density" Section.

X-Ray Computed Tomography
All scans were carried out at the Henry Moseley Xray Imaging Facility (HMXIF) using a Nikon XTH 225 microCT instrument, apart from sample OUT 18021, which was scanned with a Nikon XTH 320 Custom Bay because of its large size.The meteorites were each put inside an LDPE bag and placed into a suitable size tube with OASIS ® foam to prevent movement of the sample during the scan.In some cases, multiple meteorites of similar sizes were put in one tube with foam between the bags.Reference images were taken: a dark image with the detector shuttered, and a light image without the sample, to correct for background variations.The analytical parameters (summarized in Table 4) were chosen for each scan to give as wide a range of x-ray attenuation (and thus greyscale values) as possible and to maximize resolution, subject to keeping the overall scan times reasonable (usually under 2 h per sample).Copper filters of different thicknesses were used to remove low-energy x-rays and pre-harden the x-ray beam.A ring artifacts reduction method was applied for each scan and a beam hardening algorithm was applied to the data before reconstruction.Scans were reconstructed, using a modified Feldkamp backprojection algorithm (Feldkamp The scan parameters used varied for meteorites of different sizes and densities.The different thicknesses of copper filters (0.1-2 mm), energy (170-310 kV) and exposure time (500-2000 ms) were used to optimize contrast and reduce artifacts for each sample (Table 4).Beam hardening produced artifacts in some scans and in some cases made it difficult to define the edges of the sample.Using lower energy x-rays would give more issues with beam hardening artifacts; using higher energy x-rays would give less sensitivity to attenuation contrasts (Ketcham & Carlson, 2001).The energy levels used may give rise to thermal drift, causing small movement in the position of the beam, resulting in a blurring effect that could cause more error in defining the edge accurately (de Oliveira et al., 2014).Given the beam spot size of $5 μm, this could have noticeable effects for those scans with resolutions below $15 μm, and slight effects for resolutions below 25 μm.Ring artifacts affect the interior of the scanned volume so were not thought to cause problems with the definition of the edge of the sample.
The 3-D meteorite volume was segmented in Avizo 2019.3 (Thermo-Fisher Scientific, Waltham, MA, USA) by thresholding using user-selected parameters, first defining three materials where "Inside" is the selected voxels that are part of the meteorite, "Outside" is where selected voxels are background, and "Unassigned" is the boundary edge region where selected voxels are a mixture of meteorite and background (Figure 5b).The watershed tool in Avizo was used to automatically assign the voxels of material "Unassigned" into the other two categories by examining the difference in greyscale of the neighboring voxels to find the point of greatest contrast (Figure 5c).
Several scans were segmented two or three times, over a period of a year, to test whether user judgment of the seed regions "Inside" and "Outside" affected sample volume results; the differences were negligible (e.g., <0.01 cm 3 variation in sample volume for OUT 18019: 3.36 cm 3 ).This shows that the automated "Watershed" tool is robust at picking the same point of greatest contrast if the seed regions are reasonable.

AMetMet Measurements
The AMetMet, a closed device, contains an internal circuit that generates a magnetic field.When a sample is held close to the exterior (back) of the device, the magnetic field is disrupted.These analyses were conducted in free air away from metal tables, equipment, rings, or watches.Further details on the operation of this instrument are presented in an upcoming study by Uehara and Gattacceca (2023).The degree to which the field might be disrupted by a given sample relates to the amount of material within the sample that has been magnetized by the applied field (Gattacceca et al., 2004;Uehara & Gattacceca, 2023).A sample must be above a certain mass (0.5 g) for accurate readings, and large samples (larger than the instrument) can be measured (Gattacceca et al., 2004).
The AMetMet device used in this study is a small, handheld device that is operated by pressing the control button when the sample is held to the device.Following an initial "blank" air test, samples were placed close to the probe, disrupting the field of the internal circuit.The extent of disruption correlates to magnetic susceptibility (Logχ where χ is the ratio of the magnetization of the sample to the applied magnetic field strength) and electrical conductivity (this value is relative to an arbitrary scale used by the AMetMet device).The dial on the front of the device was used to enter the appropriate mass of the sample (from a range of pre-registered masses), to apply a sample mass correction.Measurements were made away from metal objects, surfaces, and worn rings that might also interfere with the reading.Each meteorite was measured (blank air test followed by sample test) six times and average values for the magnetic susceptibility and electrical conductivity were determined from the six measurements.Errors reported on these values in Table 4 are two standard deviations on these average values (n = 6), and the full data set is presented in Table 4.

Volume from Photogrammetry
We produced high-quality 3-D models of 18 meteorites including a range of masses ($1 to $2500 g) and sample sizes ($1 × 1 cm up to $15 × 20 cm).The models produced from several of these meteorites are shown in Figure 6, which illustrates the range of size, surface texture, and morphological complexity.Screenshots of each final model are included in Appendix B (Figures B1-B18) and the full data set can be accessed from links in the Supporting Information.
A summary of measurements performed on the suite of Antarctic meteorites is presented in Tables 3 and 4. The sample volume values produced by photogrammetry range from 0.2 AE 0.004 to 760.14 AE 15.2 cm 3 .
To investigate possible variations in volumes derived from a single measured axis, we measured along different axes of the samples to recalibrate the sample volume.As can be seen in Table 3, these recalibrated volumes are close to the original volume determinations (all except one sample are within error), with the alternative set of calibrations (ranging from within 0.51% to 2.68% of the original calibration) not systematically higher or lower.This suggests that the produced volume models are not significantly inaccurate.In future work, alternative methods to model scaling, such as the use of a volume standard during the photogrammetry process, could be a more robust approach to accurate scaling of the models.

Volume from X-Ray Computed Tomography
We performed XCT analysis on 17 of the 18 meteorites in the sample suite.Table 1 in Appendix A shows the parameters used and resolution for each meteorite scan, as well as the volumes derived from segmenting the XCT scans, ranging from 0.19 to 758.0 cm 3 .

Bulk Density
Bulk density values for each meteorite were determined using the mass of the samples and volumes determined by both photogrammetry and XCT (Table 4).Bulk density values computed by photogrammetry range from 3.13 AE 0.06 to 4.47 AE 0.09 g cm À3 .The reported errors in these data are derived from the propagated errors on volume and mass values reported above.We employed a standard approach for propagating this error using the following equation: Bulk density values computed using XCT volumes differ slightly from those computed using photogrammetry volumes, ranging from 3.24 to 4.56 g cm À3 (Table 4).We have not determined a numerical estimate of analytical uncertainty for XCT-derived values because the scans were performed with different filter materials, thicknesses, and settings with different resolutions to optimize scans for meteorite size, shape, and density.Analyses were not conducted with a reference material, meaning that an instrumental error on voxel size cannot be quantified.

Magnetic Susceptibility and Electrical Conductivity
Sixteen of the Antarctic meteorites produced reproducible readings with the AMetMet (Table 4), and the two which did not were very small (<1.5 cm, <2 g).Note that at the time of publishing, Uehara and Gattacceca (2023) recommend a minimum mass of 5 g for use of the AMetMet device.Average magnetic susceptibility ranges from Logχ values 3.42 AE 0.46 to 5.47 AE 0.07 (where χ has units 10 À9 m 3 kg À1 ).Average electrical conductivity measurements (C) range from values of 0 AE 1.01 to 4.87 AE 0.15 (a unitless value).Both magnetic susceptibility and electrical conductivity values were measured six times, and the reported error is two standard deviations on the mean.A complete data table of AMetMet measurements can be found in Table S1 of the electronic supplement.
AMetMet data are plotted in Figure 8 along with measurements from the test meteorites described in "Antarctic Meteorites" Section.Data for the test meteorites are comparable with the measurements made by CEREGE from other hot desert meteorites of various classifications.Collectively, these measurements provide fields for comparison with collected Antarctic samples.There is a range of magnetic susceptibility and electrical conductivity between different pieces of the same meteorite, as is the case for test meteorites Gao Guenie (H5 ordinary chondrite), and Campo del Cielo (IAB iron) (Figure 8 and Table 2).This might be attributable to  variation in the proportion of Fe,Ni metal and sulfides in each sample, although there is no reason to expect significant variation between samples.The range of electrical conductivity and magnetic susceptibility for pieces of NWA 869 (L3-6 ordinary chondrite) is slightly greater than of the other test samples.This may be due to the brecciated nature of the sample: brecciation has been identified previously as a source of heterogeneity of magnetic susceptibility (Rochette et al., 2003).The spread of magnetic susceptibility values of NWA 869 (Table 2) might also be attributable to the nonlinear relationship between metal content and magnetic susceptibility.Samples that contain high proportions of well-connected metal have a similar magnetic susceptibility value to those of purely metallic meteorites (Rochette et al., 2009).
The Antarctic terrestrial rocks, including a gabbro, a dolerite, and two amphibolite samples (see Table 2), contain no conductive Fe,Ni metal, but do contain a range of magnetic oxides which accounts for the range in magnetic susceptibility.
Our measurements for meteorites collected from Antarctica are generally consistent with previously classified meteorite samples, although some Antarctic meteorites analyzed here plot slightly outside the range of previous meteorite measurements (Uehara & Gattacceca, 2023).The ordinary chondrites have a range of intermediate magnetic susceptibility and electrical conductivity values.The increase in magnetic susceptibility and electrical conductivity from LL to L and H ordinary chondrite groups is consistent with the increasing metal content, as has been recognized previously (Rochette et al., 2003).Several Antarctic H chondrites have higher electrical conductivity values than literature H chondrites (Uehara & Gattacceca, 2023), suggesting that our Antarctic meteorites may have higher metal contents, or that this metal is well connected within the sample.Additionally, one L chondrite (OUT 18038) lies closer to the field of LL chondrite literature data than the L field.This L chondrite sample is very small (1.13 g) and although it produced AMetMet measurements, it is possible that it is not large enough to give a representative reading.
The literature data for carbonaceous chondrites, plotted in Figure 8, includes CV, CK, CO, and CH chondrites, which vary greatly in magnetic mineralogy.The positive correlation between magnetic susceptibility and electrical conductivity reflects the amounts of electrically conductive metal, as well as nonconductive minerals such as magnetite (Rochette et al., 2008).The CM-An carbonaceous chondrite OUT 18012 has low magnetic susceptibility and zero conductivity, consistent with a low abundance of metal and magnetic minerals.
The two mesosiderite samples (OUT 18014 and OUT 18018: MacArthur et al., 2022) have high magnetic susceptibility and electrical conductivity, consistent with a high abundance of well-connected Fe,Ni metal grains.These data lie within the range of pallasites and iron meteorites, and differ from the literature data for mesosiderites, for which there is only one previous measurement (Uehara & Gattacceca, 2023).This could be because our mesosiderites contain more metal than the mesosiderite that has been measured previously.

DISCUSSION
The purpose of the following discussion is to evaluate the data determined by the methods outlined above.We compare the results of the methods that we have employed to determine meteorite bulk density.We investigate how best to use our data on the physical properties of meteorites, to distinguish between meteorite types for classification purposes.Finally, we compare our bulk density data for the recovered meteorites with literature data and draw comparisons with proposed parent bodies for the different meteorite types.

Evaluation of Methods Used to Determine Volume and Density
A computed density is highly dependent on the accuracy of the volume and mass measurements used.Measuring sample mass is not a problem, but measurement of sample volume is more complex.Measurement of sample volume can be achieved using a variety of techniques.Each method for determining sample volume has strengths and weaknesses, in terms of accuracy of measurement, accessibility of method, and physical effects to the sample.These factors, outlined below, must be considered when considering the analysis to be performed on a meteorite sample.
Traditional Archimedean methods, which involve submerging the sample in a liquid to measure the volume by displacement, are unacceptable to researchers that study meteorites, as they would result in sample damage and contamination.The related Archimedean bead method involves placing the sample in a container with small ($40-750 μm) glass beads of known volume that behave similar to a fluid (Kiefer et al., 2012;Macke et al., 2011).The displacement of these beads thus allows for a determination of the bulk volume, although this is not a surface-bounded volume because the beads can fill fractures in the sample surface.This method has been used on a wide range of meteorite samples and is the main method that has been used to determine meteorite volumes.This method has been shown to produce reliable data for meteorites of a range of sizes.It is relatively quick and has minimal physical and chemical effects on a sample compared to submersion in liquid (Consolmagno et al., 2008).However, it has been shown to have large errors for lower volume samples (<5 to 10 cm 3 ) (Consolmagno et al., 2008;Macke et al., 2009;McCausland et al., 2011;McCausland & Flemming, 2006).It also incurs systematic errors associated with the packing behavior of the glass beads of a certain size, especially when the sample is very dense.It can be unsuitable for fragile rock samples, as contact with beads may dislodge fragments, and the sample requires extensive handling to perform the method repeatedly, including brushing beads off the sample surface (Consolmagno & Britt, 1998;Consolmagno et al., 2008;Macke et al., 2009;McCausland & Flemming, 2006).
Laser scanning of the sample exterior has been employed to develop a 3-D volumetric model of meteorite samples (e.g., Macke et al., 2015;McCausland et al., 2011).Some laser imaging methods appear to incur a systematic underestimation of volume, and consequent overestimation ($2%) of bulk density relative to literature values (determined by the bead method) for the same meteorites, although the reason for this is unclear (McCausland et al., 2011;Smith et al., 2006).A more recent study by Macke et al. (2015) was able to determine sample volume measurements with a high degree of reproducibility.Multiple high-resolution scans were within 0.001% of one another for a sample of 21.1223 cm 3 in volume.This method is particularly applicable to fragile, friable samples that might be damaged by other methods or small samples that might incur large errors using displacement methods.Laser scanning requires a highly specialized laser metrology system to operate and most do not preserve a color record of the sample exterior, although this is a secondary interest when investigating meteorite physical properties.Furthermore, McCausland et al. (2011) found that samples with large cut faces or sharp corners introduced difficulties in successful model production.Macke et al. (2015) stated that laser scanning would likely replace the bead method in future meteorite density studies.Although there are several studies that have included this method for individual sample characterization (e.g., Anderson et al., 2021;Goodrich et al., 2019;Ruzicka et al., 2020), a comparable study to Macke (2010) using this technique would be beneficial for further comparison with the photogrammetry method.
XCT has a range of applications in planetary science and is commonly employed as a tool to understand information about the internal structure and composition of an extraterrestrial sample without cutting into it (Gawronska et al., 2022;Hanna & Ketcham, 2017).Tomographic studies of meteorites have been used to make a measurement of sample volume, or of the volume of components within a sample for a range of meteorites (Ebel & Rivers, 2007;McCausland et al., 2010;Wilbur et al., 2023).The key strengths of XCT are that it does not deform the sample, and that it can be used to determine a measure of internal void space within a sample.However, it has also been shown that sample porosity below the resolution of an XCT scan can also be present (Friedrich et al., 2008;Lewis et al., 2018).Although the technique does not involve damage to the bulk sample, the x-rays do physically interact with the sample on a subatomic level (Hanna & Ketcham, 2017).This has been demonstrated to affect the natural radiation record of a meteorite, which might complicate age determinations of samples such as their terrestrial residence time (Sears et al., 2013(Sears et al., , 2016)).There is also evidence that synchrotron x-rays have effects (molecular changes such as modification, fragmentation, and racemization) on free amino acids (Moini et al., 2014), although it has been shown to have no effect on amino acid abundance (Friedrich et al., 2016).
Applications of photogrammetry in Earth Sciences have included paleontological studies (Falkingham, 2012), mapping of outcrop scale geological features (Bistacchi et al., 2015) and modeling planetary surface topography (Le Mouélic et al., 2020).Using a suite of overlapping images depicting a meteorite in a range of orientations, it is possible to generate a 3-D model of the meteorite that records its shape and appearance (e.g., Harbowo et al., 2022;Harvey et al., 2020Harvey et al., , 2021)).Photogrammetry allows for production of a detailed model of the sample exterior that can be used for a range of applications including preservation of a record of whole sample exterior and determination of sample volume.Models produced using photogrammetry have been used for morphometric studies of iron meteorites (Ashley et al., 2019) as well as for the construction of 3-D sample models such as the NASA Astromaterials 3D curation database project (www.ares.jsc.nasa.gov/astromaterials3d; Blumenfeld et al., 2019).Although photogrammetry has not previously been used to determine the volume of meteorites specifically, it has been used to estimate the volume of rock deposits (Yilmaz, 2010).The strengths of photogrammetry for determining sample volume are primarily that it is completely nondestructive and noncontaminating to the sample, because with appropriate laboratory handling, the sample is not subjected to anything other than electric lights.It is also applicable to subjects of various sizes, depending on the resolution of the contributing images.Thus, with an appropriate setup, it should allow us to determine sample volumes for samples smaller than the lower limit of the bead method.The high quality of modern cameras also means that an effective setup is affordable (e.g., using a camera < £1500) (Yilmaz, 2010) and could be implemented in many labs at a wide range of institutions.The main drawback of photogrammetry is the long computer processing time required to produce high quality models, although a practiced user can produce a complete model in several days, and multiple projects can be worked on simultaneously.There are also some limitations on what objects can be modeled, such as samples with a reflective surface (e.g., polished iron meteorites) due to issues with the effect of varying reflections on pixel matching (Merkel, 2019).Photogrammetry may also not be able to produce models of samples with complex concave voids with openings at the surface (Merkel, 2019).

Photogrammetry and XCT Comparison
We can use our data to evaluate the accuracy of sample volumes derived by the photogrammetry and computed tomography methods for 17 meteorites (Table 4).Figure 9a shows a comparison of volume measurements for the suite of Antarctic meteorites produced from photogrammetry and XCT studies, as a function of sample mass.Sixteen of the samples have photogrammetry-derived volumes that are, to varying degrees, higher than the XCT-derived volume (Table 4).For the eight largest samples (>150 g: Figure 9a), photogrammetry-derived volumes are mostly slightly higher than XCT-derived volumes and volumes produced by both methods agree to within 5.5% of each other (column labeled (c), Table 4).For HUT 18029 (811 g), the XCT volume is slightly higher than the photogrammetry volume, but the two values are within error (240 cm 3 for the XCT volume vs. 237.2AE 4.75 cm 3 for the photogrammetry volume).For each of the six smallest samples (<13 g: Figure 9a), the photogrammetry method also gives a higher volume.However, differences between the volumes calculated by each method are large, with photogrammetry-derived values an average of $17% higher (and up to 35% higher for the eucrite HUT 18035) than XCT-derived values.For three meteorites of intermediate size (13.99-33.91g: Figure 9a), the photogrammetry method also gave the higher volume.Photogrammetry and XCT-derived volume values were in better agreement (OUT 18007: 3.7%; OUT 18012: 6.5%; HUT 18030: 8.4% relative difference) than values determined for the small samples.
Figure 9b shows a comparison of bulk density values derived from photogrammetry and computed tomography volume calculations.The disparity between photogrammetry and XCT volume values propagates through the density calculation.For low-mass samples (<10 g), photogrammetry-derived bulk densities are consistently lower than XCT-derived bulk densities.For most samples >10 g, photogrammetry-derived bulk densities are also lower than those derived by XCT, although many of these are within the error of the photogrammetry derived values.
Whether the disparity between photogrammetry and XCT-derived values results from an issue with a particular method can be investigated by comparison of computed bulk density values with literature data.We use literature data from Macke (2010), which were determined and rigorously repeated using the glass bead method, for this comparison due to the range and reliability of measurements made on different groups of meteorites (Figure 9b).
In general, photogrammetry-derived meteorite bulk density values are closer to meteorite literature bulk density values than those derived by XCT.For example, the maximum reported L chondrite bulk density value is 3.86 g cm À3 (Macke, 2010).Five of six of the L chondrite bulk densities via photogrammetry fall below this value, whereas only two of six of the bulk densities calculated from XCT are below it (Figure 9b).For CMan chondrite OUT 18012, L ordinary chondrite HUT 18038, and aubrite HUT 18034, both photogrammetry and XCT-derived values for bulk density are higher than the upper value of their respective literature ranges (Figure 9b).In each of these cases, the photogrammetryderived bulk densities are closer to literature values than XCT-derived bulk densities.Based on this comparison, it appears that the XCT-derived bulk density for many samples is overestimated.We consider that the cause of the disparity between photogrammetry and XCT-derived value is predominantly due to underestimation of the volume by the XCT measurement (see below).
Except for H ordinary chondrite OUT 18005, the densities calculated via XCT for the meteorite samples with mass < 34 g are above the reported literature ranges (Macke, 2010;McCausland et al., 2011).As sample mass did not change between different techniques being conducted (i.e., a bit of the meteorite was not broken off or fell off), this means that the XCT volumes calculated for small samples are lower than expected.For low mass samples, the ratio of surface area to volume is high, so any blurring effects or artifacts in the data, causing an outer surface layer of voxels to be considered background rather than meteorite (see Figure 5), could make a significant difference to the volume calculated.XCT-derived volumes for small samples will therefore be more greatly affected by these partial volume effects (Rittweger et al., 2004).Thermal drift of the XCT instrument (see "X-Ray Computed Tomography") may also have affected the nine smallest samples causing voxel blurring effects, as they were all scanned at voxel resolutions below 15 μm, with the smallest four scanned below 10 μm (see Table 4).The greater the surface area to volume ratio of a meteorite, the greater the impact of any such issues that either method has in defining the exact edge of the sample.This is especially likely if a sample has a large surface area and relatively low mass because of an irregular shape (resulting from cracks or voids in the sample).Such issues are not always present, as McCausland et al. (2010) showed preliminary data from microcomputed tomography (μCT) analyses of meteorite fragments (>13 g) that showed good agreement with their reference bulk density values.However, they noted that smaller pieces (e.g., a 1.12 g piece of H5 ordinary chondrite Grimsby) gave slightly lower bulk densities than expected (3.27 g cm À3 compared to a reference value of 3.39 g cm À3 ).
Photogrammetry-derived bulk densities for small samples give better agreement with literature ranges than XCT-derived values, although some low mass samples (<10 g) are outside literature ranges.Both XCT and photogrammetry-derived values may be susceptible to the effects of small samples being unrepresentative of the typical bulk material expected for that meteorite type.For instance, L ordinary chondrite HUT 18038 (1.13 g) and aubrite HUT 18034 (0.81 g) both have photogrammetry-derived bulk density values outside the range of the "maximum" values for literature reported samples (Macke, 2010;McCausland et al., 2011).Such values could result from measurement of a particularly metal-rich sample, which give an unrepresentative sample bulk density, although no evidence in the CT scan for aubrite HUT 18034 suggests a higher abundance of highdensity phases than expected.Macke (2010) noted that different meteorite groups require a different mass to be considered representative, and that enstatite chondrites below $40 g had scale-dependent heterogeneities.Jarosewich (1990) showed that ordinary chondrite samples <10 g are too small to contain representative proportions of metal and silicate grains, leading to varied density measurements between samples.Small samples from meteorite falls have been shown to trend to higher porosities than larger stones from the same falls, as internal porosity may be a weak point for breakage under the stresses of atmospheric entry (Cotto-Figueroa et al., 2016;Kohout et al., 2014;Macke et al., 2018).We would expect that the maximum bulk density of a given group would be close to the grain density, with lower values derived from varying sample porosity or unrepresentative mineralogy.Alternatively, such results may also result from inaccuracies in the production of the models.In small samples (<2 cm), this could be due to the limits of the optical resolution of the camera and equipped zoom lens.
Issues with XCT relating to scan resolution and thermal drift could be mitigated by using systems equipped with thermal drift correction systems (e.g., optical magnification of the x-ray projections for higher resolution and drift compensation scans to correct thermal drift effects).XCT work performed using a synchrotron can also achieve even higher resolutions, although the science would need to be justified and synchrotron experiments are not appropriate for our purposes of developing a quick, nondestructive measure of density.Further work could be done on samples in the intermediate mass range to better constrain the lower mass (i.e., size) limit at which each method becomes ineffective, or improve capabilities for smaller samples.

Assessment of Photogrammetry-Derived Densities for the Antarctic Meteorites
The mesosiderite OUT 18018 has a high bulk density (4.47 AE 0.09 g cm À3 : see Figure 9b), consistent with the mineralogy of mesosiderites, which are composed of approximately equal proportions of silicates (such as orthopyroxene, olivine, and plagioclase) and Fe,Ni metal with associated troilite (Weisberg et al., 2006).The other mesosiderite (OUT 18014), which is suspected to be paired with OUT 18018 (MacArthur et al., 2022), has a lower density of 3.98 AE 0.8 g cm À3 .The difference between the two could be due to variation in metal abundances between the samples.Of the full literature range, 3.08-7.17g cm À3 for mesosiderite bulk density, there is one outlier which is presumably a sample dominated by metal (Macke, 2010).Most samples measured by Macke (2010) are within 3.08-5.40g cm À3 , which is comparable to the range of photogrammetryderived bulk density values for OUT 18014 and OUT 18018.
The Aubrite sample HUT 18034 has a bulk density of 4.05 AE 0.08 g cm À3 .Bulk density of aubrites determined by Macke (2010) ranges from 2.53 to 3.36 g cm À3 , as shown in Figure 9b.Aubrites are composed mostly of enstatite ($75-95 vol%), with minor proportions of plagioclase, diopside, forsterite, and Fe,Ni metal (Weisberg et al., 2006).Our value is significantly higher than the measured literature range of aubrites, which could have resulted from a high mass measurement of a small sample (0.81 g, <14 mm width) containing a relatively high proportion of dense phase or from inaccuracies in model production due to its small size.However, neither observation of the sample exterior nor investigation of the polished mount using optical microscopy, or investigation of the XCT scan, showed an unusual concentration of dense phases.
Eucrite sample HUT 18035 has a low bulk density compared to other samples in this study (2.55 AE 0.05 g cm À3 : Figure 9b).HUT 18035 has a brecciated texture containing fragments of pyroxene and plagioclase grains, as well as igneous clasts composed of granular textured pyroxene and plagioclase (MacArthur et al., 2022).Values for eucrites determined by Macke (2010) range from 2.61 to 3.12 g cm À3 , like the range of eucrite values measured by McCausland et al. (2011), ranging from 2.89 to 3.13 g cm À3 .The lower limit of the range measured by Macke (2010) is comparable to our photogrammetry-derived value.
The CM-an carbonaceous chondrite OUT 18012 has a bulk density of 3.13 AE 0.06 g cm À3 .CM chondrites are typically composed of phyllosilicates (55-90 vol%) and anhydrous silicates (e.g., olivine and pyroxene; 5-33 vol %), with minor phases (<5 vol%) such as carbonates, magnetite, Fe sulfides, and Fe,Ni metal (Howard et al., 2015;Suttle et al., 2021).Bulk density values for CM chondrites reported by Macke (2010) range from 1.88 to 2.47 g cm À3 , which is significantly lower than the value that we determined.McCausland measured bulk density of a piece and a slab of CM2 Murchison by laser scanning, yielding bulk density values of 2.39 AE 0.02 and 2.83 AE 0.18 g cm À3 , respectively, the latter of which is much closer to our value.The high bulk density of the sample may result from reduced porosity resulting from shock events or compression due to burial on the parent body.However, it is notable that the bulk density of OUT 18012 is higher even than the typical grain density of CM chondrites (average 2.75 g cm À3 ) as measured by Macke (2010), suggesting that reduced porosity is not enough to account for the differences.This higher bulk density of OUT 18012 might result from petrological differences between the CM-an classified meteorite and the meteorites measured by Macke (2010), which were all classified as CM2.The reason for the "anomalous" CM classification is that x-ray diffraction (XRD) analyses did not detect hydrous phases (such as phyllosilicates) in OUT 18012, which are usually found in CM chondrites.The interpretation of heated CMs is that they experienced heating on their parent body, leading to the replacement of low-density hydrated phases (King et al., 2021;Nakamura, 2005), which could explain the relatively high bulk density of our measurement.
The ordinary chondrites we studied are mostly within the range of literature bulk density values, as shown in Figure 9b (H chondrites: 2.51-4.21g cm À3 , L chondrites: 2.94-3.86g cm À3 , LL chondrites: 2.80-3.62g cm À3 : Macke, 2010).Measurements of ordinary chondrites by McCausland et al. (2011) predominantly sit within these literature ranges as well.However, the bulk densities of ordinary chondrites measured in this study do not follow the increasing trend we would expect because of the increasing metal and bulk Fe content from LL to L to H groups.The spread of literature values may reflect varying porosity between samples (Macke, 2010).L6 chondrite HUT 18038 (4.04 AE 0.08 g cm À3 ) is slightly above the literature range for L ordinary chondrites (Macke, 2010).This might have arisen from inaccurate model production due to its small size or from a high proportion of dense phases, but this value is not significantly outside the range.
The problems with photogrammetry-derived bulk density values for low mass (<10 g), small samples are a limitation of the photogrammetry method for measuring volume.In low mass samples that do fall within the literature range, we have lower confidence in assessing whether this value represents a genuine result.For low mass samples with bulk density values outside literature ranges (aubrite HUT 18034, the eucrite HUT 18035, and L6 ordinary chondrite HUT 18038), it is unclear whether this results from an unusually high abundance of highdensity minerals (i.e., a genuine outlier) or issues from the method.
Additionally, working with a range of more tailored camera equipment might allow us to improve our confidence in small samples.With the appropriate camera lens, it might be possible to extend this technique to very small samples (i.e., <5 mm) such as micrometeorites.In this study, models were produced by one operator, but it would also be useful to understand variability in the outcome of the workflow between different operators.For density measurements of meteorites made using laser scanning, McCausland et al. (2011) determined a coefficient of variability for values determined by three independent operators.For each sample, they determined the ratio of the standard deviation of the operators' volumes to the mean volume.Inter-operator variation was <2% for samples >4 cm 3 and 2%-4% for samples <4 cm 3 .A similar approach applied to the photogrammetry workflow would help to assess its value in comparison to other techniques.

Classification of Meteorites Using Measured Physical Properties
Magnetic susceptibility measurements of meteorites can distinguish between the main meteorite types (e.g., between iron meteorites, ordinary chondrites and achondrites) as they have varying metallic iron content, and metallic iron has a high magnetic susceptibility relative to silicate minerals (Rochette et al., 2003).However, the range of magnetic susceptibilities for meteorite classes such as carbonaceous chondrites can overlap significantly with a range of other meteorite classes, including ordinary chondrites and achondrites, because of the variability of magnetic mineralogy, including metal content among and within different carbonaceous chondrite groups (Macke, 2010).Furthermore, different groups of meteorites with overlapping ranges of metal abundances (e.g., L and LL ordinary chondrites), have overlapping ranges of magnetic susceptibility, meaning that the meteorite group cannot be definitively inferred from its magnetic susceptibility properties (Consolmagno et al., 2006).Overlapping magnetic susceptibility ranges within individual meteorite groups are attributed primarily to terrestrial weathering.It has been shown that comparison of magnetic susceptibility data for meteorite falls, which have not had time to undergo significant terrestrial weathering, allows for better distinction between meteorites of different groups than data from finds (e.g., LL, L, and H ordinary chondrites; Rochette et al., 2003).It has also been shown that for some groups of meteorites (e.g., L and H ordinary chondrites), samples of higher petrologic type have slightly higher magnetic susceptibility than those of lower petrologic type (Rochette et al., 2003).The shape and orientation of metal grains, which change with metamorphism, could be responsible for higher magnetic susceptibility values among samples with similar metal content (Rochette et al., 2003).However, this is not always the case, as LL chondrites show no correlation between petrologic type and magnetic susceptibility.Instead, magnetic susceptibility is correlated with tetrataenite (Fe 50, Ni 50 ) abundance (Rochette et al., 2003).Collectively, these factors mean that while magnetic susceptibility measurements are a useful general guide for differentiating between meteorite groups, they are of limited value for meteorite classification in isolation.
A combination of magnetic susceptibility with bulk density provides a more robust tool for distinguishing between the meteorite groups (Consolmagno et al., 2006;Terho et al., 1991).Bulk and grain density are strongly linked to the bulk meteorite abundance of iron, and the oxidation state of that iron (i.e., Fe 0 , Fe 2+ , Fe 3+ ), in the sample.Including a measure of bulk density allows for further separation between meteorite groups that contain a high proportion of dense, magnetic phases (e.g., Fe,Ni metal) and those which contain a high proportion of dense phases with low magnetic susceptibility (e.g., paramagnetic silicates and antiferromagnetic troilite: Consolmagno et al., 2008;Rochette et al., 2003).Figure 10 shows photogrammetry bulk density of the Antarctic meteorite samples plotted against magnetic susceptibility, as well as literature data for a range of different meteorite types (Macke, 2010).Most of the density and magnetic susceptibility values we determined are within the range of previously reported meteorite data.The H, L, and LL ordinary chondrites from our study span the range of magnetic susceptibility and bulk density values for ordinary chondrites in Macke's study (Macke, 2010), although the bulk density values of some of our samples are higher than the literature range, as discussed above.
Although the plot of bulk density versus magnetic susceptibility (Figure 10) allows for better distinction between meteorite groups than magnetic susceptibility alone, there is still a significant overlap between some of the meteorite classes (e.g., between carbonaceous and ordinary chondrites).Using grain density rather than bulk density has been shown to provide significantly better separation between ordinary chondrite groups due to variable sample porosity (Consolmagno et al., 2006).
The magnetic susceptibility and electrical conductivity data (Figure 8) provide another means for classifying meteorites.Ordinary chondrite groups are mostly distinct, although the fields of data from these groups do overlap somewhat.As is the case in Figure 10, data for carbonaceous chondrites overlap with the ordinary chondrite fields because of the range of metal and magnetic oxides and sulfides that they contain.The data are mostly effective at distinguishing stony achondrites from carbonaceous and ordinary chondrites, although there is some overlap (e.g., between HEDs and some chondrites).The electrical conductivity data also allow for better division between metal-bearing chondrites and iron meteorites than the bulk density or magnetic susceptibility data (Figures 8 and 10).
As mentioned, the distribution and orientation of metal within a meteorite sample can affect magnetic susceptibility measurements, meaning that magnetic susceptibility cannot always effectively differentiate between samples where there is >$50 vol% metal (Rochette et al., 2009).As a result, there is overlap in magnetic susceptibility values between samples with high metal content, such as iron meteorites and mesosiderites (Figure 10).If magnetic susceptibility is affected by the amount and distribution of metal in a sample, then similar effects might be observed in electrical conductivity measurements.Soini et al. (2020) noted that thermal conductivity of meteorite samples is inversely correlated with porosity.As thermal and electrical conductivities are both partly reliant on the distribution and connectivity of metal in the sample, similar trends might be expected in electrical conductivity data of future studies where porosity is quantified.Comparison of bulk density, magnetic susceptibility, and electrical conductivity measurements for ordinary chondrite samples of different petrologic type shows no apparent trends (Figure 11).However, we do not have a large or diverse enough sample suite at the moment to test the extents of such effects.
The combination of magnetic susceptibility and electrical conductivity provides a useful tool for meteorite classification.It is possible to make a more confident estimate of the classification of a sample, without causing damage or laboratory contamination, rather than just relying on visual inspection of a non-fusion-crusted hand specimen.A combined technique, including a measure of bulk density, such as the one employed here, is an appropriate method for doing an initial investigation of samples within a large meteorite collection prior to a more  A1 and A2).(Color figure can be viewed at wileyonlinelibrary.com) detailed set of microanalysis or isotope measurements (Joy et al., 2021).
It is important to note that while our samples all underwent Antarctic weathering, the samples studied by Macke (2010) include falls as well as meteorites that underwent Antarctic and hot desert weathering.For many of our meteorites, the weathering state of our meteorite samples is similar to that of the finds studied by Macke (2010), although the falls in that study are unweathered.We might not necessarily expect that this would have an impact on sample bulk density (Consolmagno et al., 2008).However, we might expect to see that the magnetic susceptibility of weathered samples would be lower compared to the data for falls (Rochette et al., 2003).This is true for the case of most of our mesosiderite and L ordinary chondrites, which are at the lower end of the range of the literature data (Figure 10).However, this is not the case for our H ordinary chondrites, which span most of the range of the literature data (Figure 10).To determine whether such variation is linked to weathering, it may be useful to compare the electrical conductivity of samples of varied weathering state, as electrical conductivity being reduced by the alteration of metal might be easier to identify than alteration of magnetic metal to magnetic alteration products.

Interpreting Parent Body Properties from Measured Bulk Density
Physical properties such as those measured in this study are key parameters for understanding the mineralogy and structure of bodies from which meteorites are derived (Flynn et al., 2018).In conjunction with remote sensing, properties such as bulk density, inferred macroporosity, and internal structure of airless bodies can be determined.An example of the value of understanding meteorite properties was demonstrated by NASA's Dawn mission (Ruzicka et al., 1997).Volume estimates for asteroid 4 Vesta obtained by the Dawn spacecraft allowed for an updated determination of the asteroid's bulk density.These results were anchored using laboratory-based analysis of the bulk density of howardite, eucrite, and diogenite (HED) meteorite samples that helped to develop a model of the internal structure of asteroid 4 Vesta (Ruzicka et al., 1997) and support the interpretation that 4 Vesta underwent a magma ocean phase early in its history (McSween et al., 2013;Russell et al., 2012).Likewise, conclusions drawn from NASA's Gravity Recovery and Interior Laboratory (GRAIL) mission about the density and porosity of the Moon's crust were constrained by laboratory studies of the physical properties of Apollo lunar samples and lunar meteorites (Kiefer et al., 2012;Wieczorek et al., 2012).Using this information, Wieczorek et al. (2012) were able to refine values, from a previous estimate of 50 km for the average thickness of the lunar crust to a more constrained 34 km (nearside) and 43 km (farside), helping to determine the bulk chemical composition of the Moon, which is of key importance for understanding its origin and early differentiation (Taylor & Wieczorek, 2014).
Most density measurements of solar system bodies that have been determined by remote sensing are of bulk density rather than grain density because orbital or flyby missions typically cannot measure internal voids.This means that it is important to know the range of bulk densities in meteorite samples.Depending on the type and structure of a given asteroid, the amount of porosity varies (Flynn et al., 2018).For instance, asteroids that formed as rubble piles (e.g., 162173 Ryugu and 101955 Bennu: Grott et al., 2020) might contain a considerable amount of void space in the form of microporosity and macroporosity (Flynn et al., 2018;Grott et al., 2020;Herbst et al., 2021;Watanabe et al., 2019;Yada et al., 2021).Asteroid microporosity results from void space in the rock, such as pores and fractures, and is typically inferred from meteorite porosity.Macroporosity is derived from larger voids, fractures in the structure of the body itself, or in the case of a rubble pile, the space between the constituent rocks themselves (Flynn et al., 2018).It is also important to bear in mind that the estimates of bulk density for the asteroids are for unweathered materials.Although our samples have not undergone major levels of terrestrial weathering, this process is likely responsible for some of the differences observed between terrestrially recovered meteorites and inferred asteroid bulk densities.
The two mesosiderite samples (OUT 18014 and OUT 18018) have photogrammetry-derived bulk density values of 3.98 AE 0.08 and 4.47 AE 0.09 g cm À3 , respectively.Due to the larger size of these samples and apparent similarity to previously measured mesosiderite literature values, we can be confident in the bulk density values for mesosiderites produced by our methodology.These values can be used for comparison with bulk density data for the bodies that have been suggested as candidates for the mesosiderite parent body.Based on remote spectral analysis and physical property comparison, several asteroids have been considered as potential parent bodies for the mesosiderites, including 16 Psyche, 201 Penelope, 250 Bettina, and 337 Devosa (Elkins-Tanton et al., 2020;Vernazza et al., 2009).Recent estimates for the bulk density of asteroid 16 Psyche have a wide range, from 2.54 AE 0.91 g cm À3 (Sitala & Granvik, 2019) to 7.59 AE 1.2 g cm À3 (Somenzi et al., 2010).Elkins-Tanton et al. (2020) showed that most of the estimates for Psyche's bulk density are between 3 and 5 g cm À3 .The mesosiderite bulk density range determined by Macke (2010) agrees with estimates for the bulk density of 16 Psyche.Our mesosiderite photogrammetry-derived bulk density values are also comparable to the recent bulk density asteroid value determined by the Psyche mission pre-flight study of 3.78 AE 0.34 g cm À3 (see Figure 12: Elkins-Tanton et al., 2020).It is worth noting that our bulk density values for OUT 18014 and OUT 18018 are at the lower end of the mesosiderite range measured by Macke (2010), which ranges from 3.08 to 7.17 g cm À3 .Thus, data for these samples determined using our method support an argument that asteroid 16 Psyche may be the mesosiderite parent body, based on bulk density measurements alone.
Evidence from remote reflectance spectra suggests that the surface mineralogy of Psyche contains silicate minerals with lower FeO contents than those observed in mesosiderites and may be more like the CB chondrites (Elkins-Tanton et al., 2020).However, it is unclear whether remotely observed surface composition provides indicative information about the composition of the interior of the body (Ostrowski & Bryson, 2019).Furthermore, an asteroid may have a lower density than meteorites from which it is derived, as internal macroporosity is a significant factor (Grott et al., 2020;Lupishko, 2006).Indeed, mesosiderites are commonly brecciated, and Elkins-Tanton et al. ( 2020) speculate that impact brecciation could cause high porosity.The 2022 mission will aim to clarify competing hypotheses about the composition of 16 Psyche (Elkins-Tanton et al., 2011, 2020).
Asteroid 337 Devosa is one of the asteroids suggested as a potential parent body for the mesosiderites by Vernazza et al. (2009) based on spectral comparison with the Vaca Muerta meteorite.Carry (2012) determined the bulk density of asteroid 337 Devosa to be 7.91 AE 1.65 g cm À3 .In this instance, bulk density data from our mesosiderites, and indeed the range determined by Macke (2010), are lower than this estimate (see Figure 12).Based on bulk density, it would therefore seem that asteroid 377 Devosa is too dense to be the parent body of the mesosiderite meteorite group and is more consistent with the bulk density of an iron meteorite (up to 7.97 g cm À3 for a fragment of IIAB iron Sikhote-Alin: Macke, 2010).
OUT 18012 is a CM-an carbonaceous chondrite with a bulk density of 3.13 AE 0.06 g cm À3 .Investigation of spectra from CM chondrite meteorites has indicated that they are close matches for C-type asteroids (Suttle et al., 2021).It is important to note that although CM chondrites are comparable to C-type asteroids, not all Ctype asteroids have CM chondrite compositions.There is a significant difference in bulk density between these asteroids and OUT 18012 (see Figure 12), which is also slightly higher than the 1.88-2.47g cm À3 range determined for CM meteorites studied by Macke (2010), all of which were classified as CM2.Ostrowski and Bryson (2019) noted that generally, the density of meteorite samples is an upper limit estimator for asteroids.Near-Earth, C-type asteroids include 162173 Ryugu and 101955 Bennu, the target bodies of the JAXA Hayabusa2 and NASA OSIRIS-REx missions, respectively, as well as 253 Mathilde, one of the targets of NASA's NEAR mission (Lauretta et al., 2019;Yada et al., 2021;Yeomans et al., 1997).The bulk densities calculated for these bodies are very low, with Ryugu, Bennu, and Mathilde estimated to be 1.282 AE 0.231 g cm À3 (Yada et al., 2021), 1.190 AE 0.013 g cm À3 (Lauretta et al., 2019), and 1.3 AE 0.2 g cm À3 (Yeomans et al., 1997), respectively.The low density of these asteroids has been attributed to both variable rock microporosity and macroporosity (Lauretta et al., 2019;Macke et al., 2011;Yada et al., 2021).For example, the macroporosity of Ryugu is estimated to range from $7% to 16%, which is within the possible range of a rubble pile (Grott et al., 2020;Herbst et al., 2021;Yada et al., 2021).Although Ryugu was initially predicted to be CM-like in composition, recent studies of returned samples indicate that it is more like the CI (Ivuna-like) chondrites (Greenwood et al., 2022;Yokoyama et al., 2022).Yada et al. (2021) report individual particles from Ryugu as dense as 1.8 g cm À3 , which is within the lower bounds of established data for the CM chondrites.Macke (2010) analyzed one CI chondrite, Orgueil (see Figure 12), determining a bulk density of 1.57 g cm À3 , which is closer to the 1.28 g cm À3 estimated by Yada et al. (2021).Furthermore, the high porosity, low bulk density material returned from Ryugu is not represented in terrestrial meteorite collections, perhaps due to sampling bias resulting from these types of poorly consolidated samples not likely surviving Earth atmospheric entry during the meteor entry event (Yada et al., 2021).
Bennu, which has been identified as spectrally comparable to aqueously altered CM-type meteorites, has an even lower estimate of bulk density (Hamilton et al., 2019;Lauretta et al., 2019;Zolensky et al., 2020).The disparity between CM measurements and the low value for Bennu most likely results from asteroid macroporosity (Suttle et al., 2021).The accessibility of the methods presented in this study will encourage continuing measurement of CM, CI, and other carbonaceous chondrite bulk densities, which could help to address these issues.Our photogrammetry method is particularly important for measuring bulk densities of very friable materials such as CM and CI chondrites.
Ordinary chondrites in this study have a range of bulk density values comparable to the range of H, L, and LL meteorites reported in the literature (Figures 9 and FIGURE 12.Comparison between bulk density of Antarctic meteorites and estimated bulk density for proposed parent bodies: 253 Mathilde (Yeomans et al., 1997), 433 Eros (Yeomans et al., 2000), 337 Devosa (Carry, 2012), 25143 Itokawa (Tsuchiyama et al., 2011), 4 Vesta (Russell et al., 2012), 6 Hebe (Marsset et al., 2017), 101955 Bennu (Lauretta et al., 2019), 16 Psyche (Elkins-Tanton et al., 2020), and 162173 Ryugu (Yada et al., 2021).Density measurement of returned material from asteroid Itokawa by the Hayabusa 1 mission (Nakamura et al., 2011).Literature data for other meteorites from Macke (2010).(Color figure can be viewed at wileyonlinelibrary.com) 10).The ordinary chondrites are thought to be derived from S-type asteroids based on spectral comparisons (Vernazza et al., 2015).It has been suggested that the different groups of ordinary chondrites, which have different petrological characteristics, are derived from several bodies (Vernazza et al., 2015;Yomogida & Matsui, 1984).Asteroid 6 Hebe has been suggested as a potential source of H chondrites, 433 Eros has been suggested as a possible source for L chondrites and the asteroid 25143 Itokawa has been shown to be a source of LL chondrite material (Britt et al., 2001;Nakamura et al., 2011;Vernazza et al., 2015).Estimates of bulk densities are 3.48 AE 0.64 g cm À3 for Hebe (Marsset et al., 2017) and 2.67 AE 0.03 g cm À3 for Eros (Yeomans et al., 2000).Asteroid 25143 Itokawa is estimated to have a bulk density of 1.9 AE 0.13 g cm À3 for the entire asteroid (Tsuchiyama et al., 2011).Although we can make good comparisons between meteorite data and density estimates for S-type asteroids that have been determined so far, it is important to note that there are many S-type asteroids whose densities have yet to be estimated, which might be the source of some members of the ordinary chondrite sample collection.Bulk density values derived using our method (which are generally comparable with the range of bulk density measurements presented in the literature: Figures 9 and 10) are plotted with these hypothetical parent bodies in Figure 12.Our lower density H ordinary chondrite data and much of the literature data agree with the bulk density of Hebe.Both literature L ordinary chondrite bulk density values and measurements from this study are higher than those of Eros.Additionally, LL ordinary chondrite values from literature and this study are higher than estimates for 25143 Itokawa.
As in the case of the CM chondrites, the disparity between measured meteorite values and the bulk densities estimated for Eros and Itokawa is likely a result of asteroid macroporosity.There is overlap between literature measurements of LL chondrites and the small mass of particles returned by the JAXA Hayabusa mission (Tsuchiyama et al., 2011).25143 Itokawa has been identified as a rubble pile asteroid with significant macroporosity (39 AE 6%: Tsuchiyama et al., 2011).Analysis of particles returned from 25143 Itokawa using XCT allowed for determination of an approximate bulk density value of 3.1 AE 0.2 g cm À3 (Tsuchiyama et al., 2011), which is comparable with literature data for LL ordinary chondrites (see Table 1: Macke, 2010), as well as OUT 18029 (3.42 AE 0.07 g cm À3 ).Tsuchiyama et al. (2011) computed grain density of the 25143 Itokawa particles of (3.4 g cm À3 ), which showed good agreement with measurements of LL chondrites (3.54 g cm À3 ) by Consolmagno et al. (2008).Although it is not a low-cost element of the workflow, determination of porosity using XCT to interpret the grain density of meteorites would be a useful tool for making more detailed comparisons.

CONCLUSIONS
We have shown that it is possible to make high fidelity 3-D models of meteorite samples, using photogrammetry.These can be used to compute sample volumes, from which bulk density can be determined.The photographic product of this method is also useful as a long-term curatorial sample record.Small samples (<10 g) were more difficult to model with our laboratory photogrammetry setup, and errors in the determination of their volume can have a large impact on the resulting density value.Future work using a range of camera lenses may reduce the effect of this error.For low mass samples (<10 g), we found that XCT-derived volumes produce bulk density values inconsistent with sample classifications, while photogrammetry-derived volumes were closer to the appropriate range.For larger samples, the two methods are more comparable.
It is not possible to use a single physical property (i.e., bulk density, magnetic susceptibility, or electrical conductivity) to classify a meteorite, because of variable features within a group such as porosity and the degree of interconnectivity of metal particles.On plots of two physical properties, such as bulk density versus magnetic susceptibility, and electrical conductivity versus magnetic susceptibility, meteorite groups are better defined although there are still overlaps that prevent definitive classification.Nevertheless, measurements derived from our method can be used broadly as a classification tool and to give an initial interpretation of sample mineralogy, with nondestructive methods that preserve the integrity of samples.
Bulk density is not just a useful tool for meteorite classification: It can be important for interpreting bulk density data from asteroids determined by remote sensing.We have demonstrated that data determined from our photogrammetry method can contribute to discussion of meteorite/parent body comparisons using sample bulk density.Our method is a useful new approach to measuring meteorite densities using low-cost technology that could be widely implemented to build a database of meteorite properties in institutions worldwide.
The results of this study make clear several possible avenues of future work that may be beneficial to the use of this method in the study of meteorite physical properties.First, further work could be done to understand the errors and limits associated with the photogrammetry method.This could be assessed by conducting further tests of volume standards of varying shape multiple times to see how reproducible these scans are.Another line of investigation could be for multiple operators to test for operator bias in data processing steps/decision-making.The methodology could also be improved to provide a more robust estimate of sample volume by the inclusion of a standard volume (e.g., a non-matte black centimeter scale cube) in the imaging stage.To assess the effectiveness of the photogrammetry method in producing true sample volumes, cross sample measurements could be undertaken between the photogrammetry approach and other methods such as ideal gas pycnometry or by laser scanning technique (Macke et al., 2015).Furthermore, emerging technologies such as structure light projection laser scanning (Loz et al., 2021), which permits for the production of truecolor, true-scale models, may also be a useful technique for producing 3-D models of meteorites for comparison with those produced by the photogrammetry method described here.Lastly, future work comparing bulk density with grain density might allow us to better understand the effect of porosity on our measurements, while maintaining a minimally invasive approach to analysis.3. Mark out 5°rotational intervals on a turntable (or piece of paper) on which the sample can sit without being moved side to side.4. Place the sample on the turntable in a white light box setup (meteorite samples were placed on a cut section of a sterilized LPDE plastic bag to ensure that they are not contaminated by the turntable, but this is unique for our sample case). 5. Light the sample using LumeCube lights with soft white light diffusers to distribute a soft light over the sample surface.Distribute the lights evenly around the sample such that the sample surface is not reflective.The brightness of these can be controlled remotely to reduce reflection.If these are unavailable, desk lamps may be a suitable alternative and have been used successfully to make models.
TAKING THE PHOTOS 6.Take photographs at 5°rotational intervals around the sample ($65-70 photos per orientation) in two or more orientations to ensure adequate coverage of the sample surface.Where there is concavity in the surface such as windows opened by cracked fusion crusts, photographing from multiple orientations improves the quality of the internal topography that the photogrammetry software can reproduce.7. Finally, measure samples along a principal axis for scale calibrating once the model has been produced.

Lightroom
Preliminary processing of the images in Adobe Lightroom, manually adjusting contrast, exposure, and white balance values, across the entire image suite, in order to produce an accurately colored 3-D model.

Load image suite as RAW files into Adobe
Lightroom (Version 4.1) or similar image processing software such as IrfanView (Version 4.33).9. Inspect the photo suite for any images where vibration has caused the image to blur-these can be deleted as they will interfere with successful pixel matching in later steps.10.Use sliders to adjust brightness, contrast, and white balance to best represent the appearance of the sample in real life.11.Export the processed images as .TIFF files so that they are compatible with Agisoft Metashape.

Agisoft Metashape
Agisoft Metashape Standard (V1.6.1) was used in the production of the meteorite models.The program works by matching pixels in different photographs to simulate the position of the camera relative to the sample surface, thereby producing a colored 3-D model of the sample.However, this is broken down into several stages-1.Align Photos (Step 13), 2. Build Dense Cloud (Step 15), 3. Align Chunks (Step 18), 4. Build Mesh (Step 20), and 5. Build Texture (Step 22).Each of these operations can be performed at multiple levels of precision (with ensuing computational costs to processing time).Earlier operations can be reperformed, but these will erase the results of later operations so considering the approach that you want to take is useful to avoid having to restart lengthy operations.12. Load .TIFF files into the Metashape workspace using the Import Cameras function.Place suites of photos from each sample orientation into separate groups called chunks.13.The first stage of processing, called Align Photos, performs a preliminary pixel matching analysis to generate a sparse 3-D point cloud (composed of, e.g., $80,000 points), and to compute the digital camera positions relative to the sample.Perform this for each of the separate chunks (orientations) that are loaded into Metashape.The amount of detail (and processing time) for the preliminary stage is controlled by the Accuracy dropdown box from Lowest to Highest.14.The sparse point cloud can be used to identify the pixels that are accurate to the sample surface and those that may be deleted, such as matched pixels from surrounding area (e.g., edges of the turntable).15.Once the sparse point clouds are reduced in size and verified for successful photo alignment, each chunk can be converted to a dense point cloud using the Build Dense Cloud operation.The dense point cloud (composed of >1,000,000 points) gives a good impression of how well the software has reconstructed the sample surface.Detail is again determined by a quality filter from lowest to highest (with more significant variation in processing time from 30 min to $8 h depending on quality).16.Like the sparse point cloud, areas of unsatisfactory dense point cloud can be removed from the chunk using the same tools as the sparse point cloud.Take care to position the model carefully such that in selecting an area of pixels you only delete the area you want, as selecting a circle through the model will select pixels on both sides of the 3-D shape.17 polygons that captures the sample surface texture but includes no information regarding color.
Meshes can be generated using either the dense point cloud or depth maps.In instances where there were gaps in the dense point cloud due to poor image alignment, we favored generating the mesh from depth maps, which are calculated from the theoretical camera positions.21.The mesh can be altered using the Smooth Mesh, Decimate Mesh and Refine Mesh functions, which respectively increase or decrease the polygon count (and resultant detail and file size) of the mesh.
22. Once the mesh is produced to a satisfactory level of detail, produce an exportable colored texture file for the 3-D models using the Build Texture operation.
The texture is composed of a mosaic of images derived from all the input images such that the color of matched pixels is colored according to the average color of those points identified in each picture.

Volume Calibration in 3DMax
The aim of the following steps is to scale the 3-D model of unknown size within 3-D space to represent its true size and use the built-in tools to compute a volume measurement.As such, it is critical to ensure that all operations are undertaken such that the 3-D aspect ratio of the model is preserved.

FIGURE 1 .
FIGURE 1. Renders of 3-D models produced to test the photogrammetry workflow: (a) H5 ordinary chondrite Gao Guenie, and (b) L5 ordinary chondrite NWA 869.The full models of these samples can be found in the online repository.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 3 .
FIGURE 3. Lightbox setup showing camera, turntable, and portable lights inside light tent.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 4 .
FIGURE 4. Stages of development for model generation in Agisoft Metashape of sample OUT 18021: (a) sparse point cloud, (b) dense point cloud, (c) un-textured mesh, (d) textured mesh.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 5 .
FIGURE 5. Images showing the segmentation process for HUT 18026: (a) Original 2-D slice of an XCT scan.(b) User-defined areas of "Inside" (light blue), "Outside" (dark blue), and "Unassigned" (black).(c) Final segmentation result after the Avizo Watershed automated tool assigned all the black "Unassigned" voxels from B as either "Inside" or "Outside" via point of greatest gradient contrast.(Color figure can be viewed at wileyonlinelibrary.com) Figure 7a,b show the segmented volume (blue) compared to the original scan (gray) for OUT 18021, and Figure 7c,d show a 2-D slice of HUT 18038 demonstrating how internal cracks were contained within the segmented edge boundary, to give a bulk volume inclusive of internal porosity and voids.

FIGURE 7 .
FIGURE 7. (a) Three-dimensional reconstruction of OUT 18021 from XCT data, (b) segmented volume from XCT scan of OUT 18021, (c) 2-D slice from the XCT scan of L6 ordinary chondrite HUT 18038, (d) segmented volume of meteorite in slice from XCT scan of L6 ordinary chondrite HUT 18038.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 9 .
FIGURE 9. (a) Volumes of Antarctic meteorites determined from photogrammetry and XCT, plotted as a function of sample mass.(b) Bulk densities of Antarctic meteorites computed using photogrammetry and computed tomography volumes, plotted as a function of mass.Ranges for literature meteorite data are from Macke (2010).Error bars on photogrammetry-derived bulk density are smaller than the size of the symbols.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 10 .
FIGURE 10.Bulk density computed from photogrammetry versus magnetic susceptibility (presented as Log χ where χ has units 10 À9 m 3 kg À1 ) for Antarctic meteorites in this study (solid symbols) and literature data (open symbols) from Macke (2010).yaxis error bars on bulk density are not shown because they are smaller than the symbols.x-axis error bars represent two standard deviations on the six magnetic susceptibility measurements averaged together for each meteorite (see Appendix A) (TablesA1 and A2).(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE 11 .
FIGURE 11.Comparison of magnetic susceptibility and electrical conductivity measurements for Antarctic ordinary chondrites with varying petrologic type (listed in Table4).(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE C2 .
FIGURE C2.Use import cameras operation to import images.(Color figure can be viewed at wileyonlinelibrary.com) FIGURE C4.Produce preliminary point cloud using align photos in the workflow tab.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE
FIGURE C16.Merge Chunks menu allows you to select the aligned chunks and merge their dense clouds to form one merged chunk containing both clouds.(Color figure can be viewed at wileyonlinelibrary.com)

FIGURE C20 .
FIGURE C20.Converted mesh in solid color viewing mode showing morphology.(Color figure can be viewed at wileyonlinelibrary.com) FIGURE C26.Model without texture in the 3DSMax workspace.(Color figure can be viewed at wileyonlinelibrary.com)

TABLE 2 .
Physical data for test meteorite and terrestrial samples.Full data in TableA2.
a Errors reported are 1 standard deviation on n = 3 measurements.

TABLE 3 .
Original and repeat axial measurement and volume determinations of Antarctic meteorite samples.
MacArthur et al., 2022)represent the 2% volume error discussed in "Volume from Photogrammetry" Section.aThis is the smaller 233.46 g stone of the OUT 18014 meteorite (seeMacArthur et al., 2022).

TABLE 4 .
Physical properties and measurement parameters of Antarctic meteorite samples.For OUT 18011, an XCT scan was not performed.Errors on average electrical conductivity and magnetic susceptibility results are two standard deviations.Errors on photogrammetry derived density are as described in "Bulk Density" Section.For HUT 18035 and HUT 18038, samples were too small (<2 cm diameter) to generate any result with the AMetMet device.Errors on sample mass were a This is the smaller 233.46 g stone of the OUT 18014 meteorite (see MacArthur et al., 2022).b Data sourced from the Meteoritical Bulletin (accessed 2023).

TABLE A2 .
Electrical conductivity and magnetic susceptibility measurements for test meteorite samples.