The scientific motivation lies in a field known as astrochronology, which exploits the fact that periodic variations in Earth’s orbit and axial orientation, the Milankovitch cycles, leave quasi-periodic patterns in sediments. These patterns can be used to construct geological timescales, reconstruct past daylength, track Earth system evolution, and probe the long-term dynamics of the Solar System. Astrochronology has become a central component of modern geochronology, contributing to the calibration of the Geological Timescale, the reconstruction of past climate variability, and studies of Earth–Moon distance and Solar System evolution. Improvements in the underlying methodology therefore ripple far beyond stratigraphy, touching multiple areas of geoscience and planetary science.

The fundamental challenge is that sedimentary records preserve orbital forcing only indirectly. Variability in sedimentation rates, changes in depositional environments, diagenetic overprinting, and nonlinear sedimentary responses all distort the astronomical signal. In deep-time studies beyond approximately 50 to 60 million years ago, theoretical orbital solutions become unreliable because of the chaotic evolution of the Solar System, so astrochronologists must instead use inverse methods that infer sedimentation histories and test their consistency with orbital forcing. Over the past two decades, a family of such methods has emerged, including Average Spectral Misfit and its moving-window extension, Bayesian spectral approaches that evaluate astronomical power against a noise model, time-domain techniques such as TimeOpt and time-variant sinusoidal modelling, correlation-based methods like COCO and eCOCO, and more recently AstroGeoFit, which jointly optimizes sedimentation rates with rigorous error estimation.

The Copenhagen team argues that these approaches share a weakness: sedimentation is an inherently unsteady and discrete process, shaped by short-lived hiatuses, implicit erosion, cementation, authigenic mineral formation, and recrystallization, all of which produce nonlinear time–depth relationships across all timescales. Methods that assume smoothly varying sedimentation can smooth away exactly the abrupt changes that carry geological meaning. AstroComb builds on the Bayesian spectral inversion framework introduced by Malinverno and colleagues in 2010 but extends it in three important ways. It simultaneously incorporates multiple observational channels, so independent geochemical proxies jointly constrain sedimentation-rate estimates. It permits abrupt local variations in sedimentation rate wherever the data support them. And it explicitly quantifies uncertainty by estimating full posterior probability distributions for sedimentation rates at every stratigraphic position.

Technically, the algorithm works in the frequency domain through a sliding-window probabilistic inversion. A fixed astronomical frequency appears in a stratigraphic record at a wavenumber that depends on the local sedimentation rate, so variations in that rate shift, split, or smear spectral peaks. For each depth window, AstroComb compares the observed multi-channel spectrum against a target astronomical line spectrum stretched by candidate sedimentation rates. Uniquely, the likelihood is built by mimicking the way human analysts identify spectral peaks: a sigmoid probability function rises between user-defined amplitude bounds, assigning each candidate peak location a probability of being a genuine Milankovitch peak. The likelihood of a given rate is the product of these probabilities across all target cycles, and when several elemental records are used, the combined likelihood is the product of the individual likelihoods, assuming independent noise between channels. A Bayesian posterior, here with a uniform prior over rates from 1.0 to 10.0 millimetres per thousand years, then yields probability densities of sedimentation rate at every depth, from which floating geological time and its variance are accumulated layer by layer.

The team validated the method on synthetic data designed as a severe test: sinusoidal Cambrian Milankovitch forcing overlain by 200 percent red noise, with a reference sedimentation model containing abrupt discontinuities of about a factor of three at depths of 70, 74, 81, and 83 metres. AstroComb recovered the true sedimentation rates with what the authors describe as excellent fidelity given the noise level, capturing the abrupt changes without smoothing them, although the timing of each change carries uncertainty. The test also exposed honest limitations: noise, model variability, window size, and the nonlinearity of the problem introduce artifacts, and a given cycle can be assigned to different target periods, producing multimodal probability distributions at some depths. Crucially, the highest probability remained associated with the true reference model, and when the algorithm was fed pure noise, it produced flat posterior distributions and low information content, demonstrating stability against spurious solutions.

For the real-world application, the researchers analysed X-ray fluorescence data from a roughly 21-metre interval of the Fågeltofta-2 drill core through the Alum Shale Formation, collected with an ITRAX core scanner at a 0.2-millimetre scanning resolution and 7 seconds per scan, yielding concentration records for 22 elements. Aluminium was selected as the primary single-channel input because of its robustness as a proxy for detrital input and its strong spectral structure in earlier work. Running AstroComb with a 3.3-metre window stepped every 0.2 millimetres, the team found that the aluminium record above 76 metres depth contains a stronger Milankovitch signal than deeper intervals, corresponding to an approximately fourfold improvement in the fit between data and target spectrum. Non-zero information content throughout the core suggests that Milankovitch cyclicity is present across the section, while sedimentation rates vary along it. The resulting floating timescale was anchored at a depth of 86.005 metres, corresponding to an age of 496.595 million years consistent with previous studies.

The multi-channel demonstration combined sulphur, aluminium, and manganese simultaneously. The authors stress that elemental channels must provide independent information: XRF elements often covary because they share mineral hosts, are affected by dilution, or are linked mathematically, and such covariance can create a misleading impression of increased certainty rather than reflecting genuine independent signals of orbital forcing. Aluminium, sulphur, and manganese meet the criteria in the Alum Shale, and although aluminium and sulphur are often anti-correlated, earlier studies show that each element records Milankovitch periodicities in intervals where the other does not, indicating that their orbital signals are not merely products of dilution. The three-channel analysis delivered a clear improvement in spectral resolution, with high probabilities increasing, low probabilities decreasing, and several secondary peaks suggesting alternative solutions reduced or absent. The information content rose by roughly an order of magnitude compared with the single-channel run.

The authors are candid about limitations. The assumed orbital target periods are model-based; for the upper Cambrian around 497 million years ago, the target includes a 405-thousand-year long eccentricity cycle, short eccentricity terms of 98.9 and 123.9 thousand years, obliquity at 32.6 thousand years, and several precession-related periods, all encoded as four sinusoids without amplitudes, since the climate response modifies amplitudes in unknown ways. Potential shifts in Earth’s rotational parameters, for example from tidal dissipation, could introduce biases. The current version computes time independently for each layer using maximum a posteriori estimates, so where two sedimentation rates are roughly equally probable, as in parts of the 74-to-78-metre interval where values near 3.2 and 9.8 millimetres per thousand years compete, expert knowledge can be added as a non-uniform prior to guide the choice. Milankovitch cycles are quasi-periodic but not sinusoidal, and nonlinear sedimentary systems can produce sum-and-difference frequencies that the algorithm ignores. The signal-to-noise ratio is generally very low, a problem the team likens to finding a needle in a haystack, and AstroComb is explicitly designed to quantify the analysis probabilistically and thereby avoid detrimental human bias.

Beyond the technical achievement, the researchers see broader implications. AstroComb operates as an unsupervised approach and detected abrupt, rather than smooth, changes in sedimentation rate in the Fågeltofta-2 core, potentially guiding sedimentologists in reconstructing pulses of deposition from geological archives. Notably, the sedimentation-rate shifts identified in the Alum Shale Formation, including at the Billegrav-2 and Fågeltofta sites, are correlated with relative sea-level changes obtained from sequence stratigraphy, opening prospects for understanding sea-level dynamics in deep time. Because sedimentation rate is a first-order control on organic matter preservation, the authors suggest the tool may serve sedimentologists as a new relative chronometer. The exact program version, input data, and scripts are publicly available under a GNU General Public License, written in MATLAB, offering the community a reproducible platform for exploring the causes and consequences of climatic, geochemical, and evolutionary change in Earth’s early history.

Subject of Research: AstroComb(v.1.0): non-linear, multi-channel, probabilistic cyclostratigraphic analysis

Article Title: AstroComb(v.1.0): non-linear, multi-channel, probabilistic cyclostratigraphic analysis

Article References: Fernandes, I., Mosegaard, K., Sørensen, A. L., Youssof, M., Thibault, N., & Dahl, T. W. (2026). AstroComb(v.1.0): non-linear, multi-channel, probabilistic cyclostratigraphic analysis. Geoscientific Model Development, 19(20), 10143-10154. https://doi.org/10.5194/gmd-19-10143-2026

Image Credits: AI Generated

DOI: 10.5194/gmd-19-10143-2026

Keywords: AstroComb, non-linear, multi-channel, probabilistic, cyclostratigraphic, analysis, scientific research