Previous research (Kiani Shahvandi et al. 2024a) has demonstrated that so-called barystatic processes—namely, the redistribution of ocean mass resulting from the melting of polar ice sheets and mountain glaciers, alongside shifts in terrestrial hydrology—exert a measurable influence on the Earth’s rotational dynamics by contributing to a gradual deceleration of its spin. This phenomenon arises because rising sea levels, particularly in equatorial regions, subtly alter the planet’s geometry, rendering it more oblate in shape. As a consequence of this redistribution of mass toward the equator, the Earth’s moment of inertia increases. In accordance with the principle of conservation of angular momentum, this increase necessitates a corresponding reduction in rotational velocity, thereby slowing the planet’s rotation. A deceleration in the Earth’s rotation inherently implies that the time required for a full axial rotation—one complete revolution—becomes slightly longer, resulting in an incremental increase in the length of a day. The referenced study provides a detailed analysis of variations in the length of day associated with climate change throughout the 20th century, noting that these changes have exhibited an acceleration as the 21st century began. Although the magnitude of this effect remains extremely small—on the order of approximately one and a third milliseconds per century—and is therefore imperceptible to human experience, it holds significant implications for domains requiring exceptional temporal precision, such as high-accuracy timekeeping systems and space-based navigation. Importantly, projections indicate that if current trajectories of climate change persist without meaningful mitigation, the rate of this rotational slowdown—and the associated increase in day length—could more than double by the end of the century. At that point, the climatic processes will dominate the variations in the length of day, overtaking the influence of lunar tidal friction.

A key unresolved question, however, has been whether the observed increase in the length of day during the 21st century represents an anomalous, unprecedented development, or whether it should instead be interpreted as part of a broader spectrum of natural variability operating over much longer climatic timescales. In other words, the extent to which contemporary changes deviate from historical patterns—or simply reflect the continuation of long-standing cycles—has remained insufficiently constrained, leaving open an important gap in the interpretation of recent observational trends. To address this limitation, a subsequent investigation by Kiani Shahvandi et al. (2024b) significantly expanded the temporal scope of available estimates of climate-induced variations in the length of day, extending the record back to approximately 720 BC. By reconstructing these variations over nearly three millennia, the study provided a much broader historical context within which modern changes can be evaluated. Their findings indicate that the climate-driven component of length of day variability remains relatively modest in magnitude, typically exhibiting amplitudes on the order of approximately one millisecond. While small in absolute terms, this signal is nonetheless detectable within high-resolution geophysical records and thus warrants careful consideration in precise temporal analyses. Importantly, the authors further identified the presence of structured variability on centennial and millennial timescales, characterized by fluctuations that exhibit increasing amplitude when viewed over longer, quasi-periodic cycles. These patterns suggest that climate-induced influences on Earth’s rotational behavior are not purely stochastic, but instead may be modulated by underlying processes that operate over extended temporal horizons. At the same time, the study underscores a critical source of uncertainty: over the past roughly three thousand years, global sea levels have remained comparatively stable, limiting the extent to which large-scale barystatic processes could have amplified rotational effects. As a consequence, it remains unclear whether the influence of contemporary climate change on length of day variations is substantially greater than historical precedents would suggest, or whether current observations still fall within the bounds of natural variability once these long-term cycles are fully accounted for. This ambiguity highlights the need for continued refinement of paleoclimatic reconstructions and Earth system models in order to more definitively resolve the relative contributions of anthropogenic forcing versus intrinsic climatic variability.

In a recent study, Kiani Shahvandi & Soja (2026) undertake a comprehensive reconstruction of climate-induced variations in the length of day, extending the temporal record back to the Late Pliocene, approximately 3.6 million years ago. This ambitious effort substantially broadens the observational horizon beyond previously available reconstructions, enabling a far more rigorous assessment of long-term variability in Earth’s rotational dynamics. To achieve this, the authors integrate outputs from advanced climate models with empirical evidence derived from paleoclimate proxies, specifically fossil records of benthic foraminifera and coral reef systems, both of which serve as well-established indicators of past oceanographic and climatic conditions. Recognizing the inherent uncertainties and noise associated with such proxy-based reconstructions, the study introduces a novel probabilistic machine learning framework designed to improve inference reliability. This approach, referred to as physics-informed diffusion models, explicitly incorporates known physical constraints into the learning process, thereby enhancing the consistency and interpretability of the reconstructed signals. By embedding physical principles within a probabilistic modeling architecture, the methodology allows for a more robust treatment of uncertainty while preserving fidelity to established geophysical relationships. The resulting reconstruction of length of day variability exhibits a high degree of internal consistency across multiple independent climate models and proxy datasets, providing strong evidence for the robustness of the inferred temporal patterns. This convergence across distinct data sources and modeling approaches lends considerable credibility to the study’s conclusions, reducing the likelihood that the observed signals are artifacts of any single dataset or methodological assumption. On the basis of this extended and methodologically rigorous reconstruction, the authors demonstrate that the contemporary rate of increase in the length of day is exceptionally high when viewed in the context of the past 3.6 million years. Such elevated rates appear to be largely unprecedented within the reconstructed record, indicating a marked deviation from long-term natural variability. Consequently, the study advances the argument that the current acceleration in length of day changes is not merely a continuation of intrinsic climatic cycles, but is instead primarily driven by anthropogenic climate change, reflecting the growing influence of human-induced alterations to the Earth system.

References

Kiani Shahvandi, M., Adhikari, S., Dumberry, M., Mishra, S., Soja, B.  (2024a). The increasingly dominant role of climate change on length of day variations. Proceedings of the National Academy of Sciences, 121:e2406930121, https://doi.org/10.1073/pnas.2406930121

Kiani Shahvandi, M., Noir, J., Mishra, S., Soja, B.  (2024b). Length of day variations explained in a Bayesian framework. Geophysical research Letters, 51: e2024GL111148, https://doi.org/10.1029/2024GL111148

Kiani Shahvandi, M., Soja, B.  (2026). Climate-induced length of day variations since the Late Pliocene. Journal of Geophysical research: Solid Earth, 131: e2025JB032161, https://doi.org/10.1029/2025JB032161