Automated Quantitative Surface Grain Density and Internal Macroporosity Retrieval for Main-Belt and Near-Earth Asteroids

Determining the surface mineralogy, bulk grain density, and internal structure of minor bodies is fundamental to understanding solar system formation, parent body differentiation, and planetary defense hazards. However, deriving physical properties from disk-integrated visible and near-infrared (VNIR; 0.4–2.5 μm) reflectance spectra remains severely hindered by non-linear multiple scattering, space weathering, photometric phase reddening, and mineral endmember collinearity. We present an automated computational framework for spectral unmixing, grain density inversion, and macroporosity retrieval. The engine converts directional-hemispherical reflectance to Hapke Single Scattering Albedo (SSA) space, nor-malizes phase reddening via an albedo-dependent sigmoidal model, and removes NEATM thermal infrared contamination. Non-negative mineral abundances are optimized using Sequential Least Squares Programming (SLSQP) bounded by Pseudo-Huber loss and an adaptive L1 LASSO sparsity penalty. Automatically selecting between Stony/Metallic and Primitive/Aqueous endmember libraries using the Akaike Information Criterion (AIC), surface cross-sectional area fractions are mapped to mass fractions to compute crustal grain densities (ρgrain). Combining derived grain densities with validated bulk density measurements (ρbulk) yields internal macroporosities (Φ). Uncertainties are propagated through 150 Monte Carlo realizations using a Lag-1 Autoregressive AR(1) noise model (φ = 0.65) to account for inter-channel correlations. Applied across 266 asteroid spectra (139 meeting quality criteria), the pipeline achieved a mean reduced chi-squared fit of χ 2 red = 0.961 (0.983 for Stony targets; 0.834 for Primitive targets). Stony asteroids yielded a mean grain density of ρ ̄grain = 4.11 ± 0.17 g/cm3, whereas featureless carbonaceous targets converged on a baseline prior (ρ ̄grain = 2.358 ± 0.002 g/cm3). Benchmark retrievals against spacecraft and laboratory reference targets—including 4 Vesta (ρgrain = 4.10 ± 0.05 g/cm3, Φ = 15.5%), 433 Eros (ρgrain = 4.49 ± 0.11 g/cm3, Φ = 34.1%), and 24 Themis (ρgrain = 2.36 ± 0.005 g/cm3, Φ = 23.2%)—demonstrate that this automated approach reliably bridges remote sensing observations and in-situ planetary exploration.