arXiv subjects

Michael K. Barker

Publications and source records attributed to Michael K. Barker.

11 recordsLinked to original sources

arXiv2026

SomBench: Benchmark Dataset for Advancing Machine Learning in Lunar Science

Lunar orbital missions, such as Lunar Reconnaissance Orbiter, Kaguya/SELENE, Gravity Recovery and Interior Laboratory, and Lunar Prospector, among others, provide rich multi-instrument observations, but their heterogeneity in sampling, projection, and conventions limits reproducible machine learning (ML). We introduce SomBench, a unified, spatially-aligned, ML-ready lunar dataset aggregating 30+ co-registered layers from ten instruments across four missions, spanning 1 meter to 20 kilometer/pixel and covering 82 degree latitude in 90 Lunar Transverse Mercator zones with two polar stereographic caps. An image-anchored tiling pipeline yields pretraining-ready multimodal tile views with leakage-safe splits, distributed as netCDF with Parquet catalogs. An application benchmark suite spans impact processes, volcanic history, and polar volatiles. Baseline experiments with ResNet-50 and SwinV2-B models confirm that each benchmark task is learnable from the released inputs, establishing reference points for future model development.

cs.CV↗

arXiv2026

Multimodal-Multiresolution Foundation Model for Lunar Remote Sensing

We present a multimodal foundation model for lunar remote sensing, pretrained from scratch on SomBench, a geographically partitioned corpus of nearly two million co-registered tile bundles spanning 11 modalities at two spatial scales (1 m/pixel and 100 m/pixel). The model adapts the TerraMind masked-token architecture with two lunar-specific extensions: acquisition geometry is provided as explicit context, and meter- and hundred-meter-scale tiles are trained jointly so that a single set of weights covers both resolutions. FlexiViT patch embeddings allow adaptation to different patch sizes without retraining, while modality-wise inputs enable flexible multimodal fine-tuning. Qualitative generation experiments suggest the model learns meaningful cross-modal correspondences, including terrain derivatives from elevation and illumination-consistent reflectance from geometry. We evaluate on four benchmarks: crater detection at WAC and NAC scales, irregular mare patch (IMP) segmentation, and polar ice prospectivity regression. Across tasks, the pretrained model matches or outperforms ImageNet-pretrained baselines and an architecturally identical random-init control. On multimodal ice prospectivity regression, pretrained variants achieve the best results, while the random-init model outperforms most baselines, suggesting gains arise from both the architecture and pretraining. Label efficiency is notable for WAC crater detection, where the pretrained model trained on 50% of the data exceeds the strongest ImageNet baseline trained on the full dataset. Among adaptation strategies, LoRA matches or surpasses full fine-tuning on crater detection and IMP segmentation while using far fewer trainable parameters, whereas full fine-tuning performs best for ice prospectivity regression. We release the pretrained checkpoint, benchmark datasets, and fine-tuning code to support reproducible lunar AI research.

cs.CV↗

arXiv2026

Improving Lunar Topography with Deep Learning Schrödinger Bridges

Increasing the resolution of planetary topography models can enable a better understanding of surface processes and geomorphology; however, existing analytical super-resolution methods are expensive and difficult to apply at large scales. Generative models provide the tools to learn complex relationships within data and can be applied at scale due to hardware accelerators and parallelization. We present a diffusion-based Schrödinger Bridge (SB) generative modeling approach for lunar topography super-resolution, connecting the distribution of low-resolution topography to that of high-resolution topography, incorporating physically-constraining optical imagery. Our approach is inspired by existing Shape-from-Shading methods, which improve a priori low-resolution topography by using optical images at the target resolution. We train SBs on a novel dataset of rendered lunar topography, emulating optical imagery from the Lunar Reconnaissance Orbiter Narrow Angle Camera. The result is a flexible approach for topography super-resolution which can provide pixel-level uncertainties in the reconstruction.

cs.CV↗

arXiv2014

The Nature and Origin of Substructure in the Outskirts of M31 -- II. Detailed Star Formation Histories

While wide-field surveys of M31 have revealed much substructure at large radii, understanding the nature and origin of this material is not straightforward from morphology alone. Using deep HST/ACS data, we have derived further constraints in the form of quantitative star formation histories (SFHs) for 14 inner halo fields which sample diverse substructures. In agreement with our previous analysis of colour-magnitude diagram morphologies, we find the resultant behaviours can be broadly separated into two categories. The SFHs of 'disc-like' fields indicate that most of their mass has formed since z~1, with one quarter of the mass formed in the last 5 Gyr. We find 'stream-like' fields to be on average 1.5 Gyr older, with <10 percent of their stellar mass formed within the last 5 Gyr. These fields are also characterised by an age--metallicity relation showing rapid chemical enrichment to solar metallicity by z=1, suggestive of an early-type progenitor. We confirm a significant burst of star formation 2 Gyr ago, discovered in our previous work, in all the fields studied here. The presence of these young stars in our most remote fields suggests that they have not formed in situ but have been kicked-out from through disc heating in the recent past.

astro-ph.GA↗

arXiv2012

A Deep, Wide-Field Study of Holmberg II with Suprime-Cam: Evidence for Ram Pressure Stripping

We present a deep, wide-field optical study of the M81 group dwarf galaxy Holmberg II (HoII) based on Subaru/Suprime-Cam imaging. Individual stars are resolved down to I~25.2, i.e. about 1.5 mag below the tip of the red giant branch (RGB). We use resolved star counts in the outskirts of the galaxy to measure the radial surface brightness profile down to μ_V~32 mag arcsec^-2, from which we determine a projected exponential scalelength of 0.70'+-0.01' (i.e. 0.69+-0.01 kpc). The composite profile, ranging from the cored centre out to R=7', is best fit by an EFF profile which gives a half-light radius of 1.41'+-0.04' (i.e. 1.39+-0.04 kpc), and an absolute magnitude M_V=-16.3. The low surface-brightness stellar component of HoII is regular and symmetric and has an extent much smaller than the vast HI cloud in which it is embedded. We compare the spatial distribution of the young, intermediate age, and old stellar populations, and find that the old RGB stars are significantly more centrally concentrated than the young stellar populations, contrary to what is observed in most dwarf galaxies of the Local Universe. We discuss these properties in the context of the comet-like distribution of HI gas around HoII, and argue for the presence of a hot intragroup medium in the vicinity of HoII to explain the contrasting morphologies of the gas and stars.

astro-ph.GA↗

arXiv2011

The Star Formation History and Dust Content in the Far Outer Disc of M31

We present a detailed analysis of two fields located 26 kpc (~5 scalelengths) from the centre of M31. One field samples the major axis populations--the Outer Disc field--while the other is offset by ~18' and samples the Warp in the stellar disc. The CMDs based on HST/ACS imaging reach old main-sequence turn-offs (~12.5 Gyr). We apply the CMD-fitting technique to the Warp field to reconstruct the star formation history (SFH). We find that after undergoing roughly constant SF until about 4.5 Gyr ago, there was a rapid decline in activity and then a ~1.5 Gyr lull, followed by a strong burst lasting 1.5 Gyr and responsible for 25% of the total stellar mass in this field. This burst appears to be accompanied by a decline in metallicity which could be a signature of the inflow of metal-poor gas. The onset of the burst (~3 Gyr ago) corresponds to the last close passage of M31 and M33 as predicted by detailed N-body modelling, and may have been triggered by this event. We reprocess the deep M33 outer disc field data of Barker et al. (2011) in order to compare consistently-derived SFHs. This reveals a similar duration burst that is exactly coeval with that seen in the M31 Warp field, lending further support to the interaction hypothesis. The complex SFHs and the smoothly-varying age-metallicity relations suggest that the stellar populations observed in the far outer discs of both galaxies have largely formed in situ rather than migrated from smaller galactocentric radii. The strong differential reddening affecting the CMD of the Outer Disc field prevents derivation of the SFH. Instead, we quantify this reddening and find that the fine-scale distribution of dust precisely follows that of the HI gas. This indicates that the outer HI disc of M31 contains a substantial amount of dust and therefore suggests significant metal enrichment in these parts, consistent with inferences from our CMD analysis.

astro-ph.GA↗

arXiv2008

The Stellar Populations of M33's Outer Regions IV: Inflow History and Chemical Evolution

We have modelled the observed color-magnitude diagram (CMD) at one location in M33's outskirts under the framework of a simple chemical evolution scenario which adopts instantaneous and delayed recycling for the nucleosynthetic products of Type II and Ia supernovae. In this scenario, interstellar gas forms stars at a rate modulated by the Kennicutt-Schmidt relation and gas outflow occurs at a rate proportional to the star formation rate (SFR). With this approach, we put broad constraints on the role of gas flows during this region's evolution and compare its [alpha/Fe] vs. [Fe/H] relation with that of other Local Group systems. We find that models with gas inflow are significantly better than the closed box model at reproducing the observed distribution of stars in the CMD. The best models have a majority of gas inflow taking place in the last 7 Gyr, and relatively little in the last 3 Gyr. These models predict most stars in this region to have [alpha/Fe] ratios lower than the bulk of the Milky Way's halo. The predictions for the present-day SFR, gas mass, and oxygen abundance compare favorably to independent empirical estimates. Our results paint a picture in which M33's outer disc formed from the protracted inflow of gas over several Gyr with at least half of the total inflow occurring since z ~ 1.

astro-ph↗

arXiv2006

The Stellar Populations of M33's Outer Regions II: Deep ACS Imaging

Studying the stellar populations in the outskirts of spiral galaxies can provide important constraints on their structure, formation, and evolution. To that end, we present VI photometry obtained with the Advanced Camera for Surveys for three fields located ~ 20' - 30' in projected distance southeast of M33's nucleus (corresponding to ~ 4 - 6 visual scale lengths or ~ 9 - 13 kpc in deprojected radius). The color-magnitude diagrams reveal a mixed stellar population whose youngest constituents have ages no greater than ~ 100 Myr and whose oldest members have ages of at least several Gyr. The presence of stars as massive as 3 - 5 Msun is consistent with global star formation thresholds in disk galaxies but could argue for a threshold in M33 that is on the low end of observational and theoretical expectations. The metallicity gradient as inferred by comparing the observed red giant branch (RGB) to the Galactic globular clusters is consistent with M33's inner disk gradient traced by several other studies. The surface density of RGB stars drops off exponentially with a radial scale length of 4.7' +/- 0.1'. The scale length increases with age in a manner similar to the vertical scale height of several nearby late-type spirals. Based on the metallicity gradient, density gradient, and mixed nature of the stellar populations, we conclude these fields are dominated by a disk population although we cannot rule out the presence of a small halo component.

astro-ph↗

arXiv2004

The Stellar Populations in the Outer Regions of M33. I. Metallicity Distribution Function

We present deep CCD photometry in the VI passbands using the WIYN 3.5m telescope of a field located approximately 20' southeast of the center of M33; this field includes the region studied by Mould & Kristian in their 1986 paper. The color-magnitude diagram (CMD) extends to I~25 and shows a prominent red giant branch (RGB), along with significant numbers of asymptotic giant branch and young main sequence stars. The red clump of core helium burning stars is also discernable near the limit of our CMD. The I-band apparent magnitude of the red giant branch tip implies a distance modulus of (m-M)_I = 24.77 +/- 0.06, which combined with an adopted reddening of E(V-I)=0.06 +/- 0.02 yields an absolute modulus of (m-M)_0 = 24.69 +/- 0.07 (867 +/- 28 kpc) for M33. Over the range of deprojected radii covered by our field (~8.5 to ~12.5 kpc), we find a significant age gradient with an upper limit of ~1 Gyr (~0.25 Gyr/kpc). Comparison of the RGB photometry to empirical giant branch sequences for Galactic globulars allows us to use the dereddened color of these stars to construct a metallicity distribution function (MDF). The primary peak in the MDF is at a metallicity of [Fe/H] ~ -1.0 with a tail to lower abundances. The peak does show radial variation with a slope of d[Fe/H]/dR_{deproj} = -0.06 +/- 0.01 dex/kpc. This gradient is consistent with the variation seen in the inner disk regions of M33. As such, we conclude that the vast majority of stars in this field belong to the disk of M33, not the halo as previously thought.

astro-ph↗

arXiv2004

Variations in Star Formation History and the Red Giant Branch Tip

We examine the reliability of the tip of the red giant branch (TRGB) as a distance indicator for stellar populations with different star formation histories (SFHs) when photometric errors and completeness corrections at the TRGB are small. In general, the TRGB-distance method is insensitive to the shape of the SFH except when it produces a stellar population with a significant component undergoing the red giant branch phase transition. The I-band absolute magnitude of the TRGB for the middle and late stages of this transition (~1.3-1.7 Gyr) is several tenths of a magnitude fainter than the canonical value of M_I ~ -4.0. If more than 30% of all stars formed over the lifetime of the Universe are formed at these ages, then the distance could be overestimated by 10-25%. Similarly, the TRGB-distance method is insensitive to the metallicity distribution of stars formed except when the average metallicity is greater than <[Fe/H]> = -0.3. If more than ~70% of all stars formed have [Fe/H] > -0.3, the distance could be overestimated by ~10-45%. We find that two observable quantities, the height of the discontinuity in the luminosity function at the TRGB and the median (V-I)_0 at M_I = -3.5 can be used to test if the aforementioned age and metallicity conditions are met.

astro-ph↗

arXiv2002

The Cosmic Radio and Infrared Backgrounds Connection

We use the radio-infrared (IR) flux correlation bet ween star-forming galaxies in the local universe to derive a simple analytical expression between the intensity of the IR background and the brightness t emperature of the radio background. This relation i s insensitive to the star formation history of the galaxies that produce the cosmic IR background (CIB). We use the observed CIB intensity to constrain t he cosmic star formation history, and the relation between the CIB and the cosmic radio background (CR B) to constrain the relative contribution of star-f orming galaxies to the CRB. Current limits on the C IB intensity predict a 178 MHz brightness temperature of ~18+-9 K, about half of the 37+-8 K inferred for an isotropic radio component. This suggests that s tar-forming galaxies and AGN contribute about equal ly to the CRB intensity at that frequency.

astro-ph↗