Introduction

The late Paleozoic (~360 to 250 Ma) was a transformative period in Earth’s history, marked by extensive tectonic reorganization, the formation of the supercontinent Pangaea, widespread glaciation, and significant shifts in global climate and sea levels1,2. These processes not only reshaped Earth’s surface but also played pivotal roles in driving carbon cycling and altering biological productivity3,4,5,6. Notably, this era witnessed the formation of extensive organic-rich shales and coal deposits7,8, which served as a major mechanism for carbon sequestration, contributing to long-term reductions in atmospheric CO2 and potentially driving global cooling9,10,11. These deposits today represent a substantial portion of Earth’s fossil fuel reserves, underpinning significant global energy resources. Understanding the driving forces behind climate and carbon cycle dynamics during this period is therefore essential for reconstructing how the Earth system responded and for appreciating the origins and distribution of key fossil resources.

Over the past few decades, substantial progress has been made in elucidating the impact of tectonic processes on global climate, including subduction zone activity, mid-ocean ridge expansion, and large igneous province eruptions12,13. These tectonic drivers influenced atmospheric CO2 levels through processes such as volcanic outgassing, silicate weathering, and the reorganization of oceanic circulation14,15. By altering the distribution of continents and ocean basins, tectonics further modulated coupled ocean–atmosphere dynamics, thereby affecting regional and global climate systems16. Simultaneously, the relatively stable components of Earth’s astronomical forcing, particularly eccentricity, have long been fundamental to understanding past climate variations17, as they modulate solar radiation received at Earth’s surface and profoundly influence global climate states18,19. Under relatively stable conditions, these orbital cycles are expected to produce distinct and continuous sedimentary records, providing reliable proxies for reconstructing past climates20. However, recent spectral analyses in cyclostratigraphy have frequently demonstrated substantial variations in the clarity and continuity of astronomical signals within sedimentary archives, raising critical questions regarding the factors that control orbital signal stability21. Specifically, the processes responsible for obscuring or disrupting these astronomical imprints remain inadequately understood, hindering our ability to interpret sedimentary records confidently and accurately reconstruct past climate dynamics.

Although previous studies have debated the relative impacts of tectonic processes and astronomical forcing on late Paleozoic climate variability22,23,24,25, most have assessed these factors in isolation, thereby overlooking their potential interactions and combined influence on climate system dynamics. To address this limitation, we adopt an integrative framework that combines geological datasets with numerical Earth system modeling. Specifically, we incorporate full-plate tectonic reconstructions, geochemical proxies such as strontium isotopes and detrital zircon geochronology, and cyclostratigraphic analyses. These are integrated with climate simulations using the Community Earth System Model (CESM1.2.2) and carbon cycle modeling based on the Long-term Ocean-Atmosphere-Sediment CArbon cycle Reservoir (LOSCAR) model. This combined approach enables a systematic evaluation of how tectonic processes influence the sensitivity of the climate system to astronomical forcing and affect the preservation of orbital signals in sedimentary archives. By explicitly quantifying the interactions between tectonic and orbital controls, our approach offers a distinct perspective on the mechanisms underlying climate state variability and organic carbon burial during one of Earth’s most dynamic geological intervals.

Results and Discussion

Tectonic phases in the late Paleozoic

Tectonic activity during the late Paleozoic profoundly impacted Earth’s climate system, shaping the geological landscape2,13 while simultaneously influencing polar ice sheet dynamics and global sea-level changes26,27. To characterize its temporal evolution, we compiled multiple complementary datasets, including subduction zone and mid-ocean ridge (MOR) lengths, surface production rates (km2 Myr−1), global detrital zircon age densities, marine strontium isotopic ratios (87Sr/86Sr), atmospheric pCO2 reconstructions, and LOSCAR-based carbon cycle simulations (Fig. 1). These records collectively constrain patterns of lithospheric recycling, continental crustal growth, mantle CO2 degassing, and their climate feedbacks across the Carboniferous and Permian.

Fig. 1: Late Paleozoic tectonic activity, climate, and sea-level changes.
figure 1

a The glaciation history with grounded ice and possible ice coverage across major regions1. b The partial pressure of atmospheric CO2 (pCO2) reconstructions compiled from multiple geochemical proxies (purple crosses)9,71,91 with a 35% locally weighted scatterplot smoothing (LOWESS) trend line. c Long-term and short-term sea-level changes relative to present-day levels85. d Lengths of mid-ocean ridge (MOR) and subduction zones13. e The intensity of plate tectonic activity through subduction and MOR dynamics13. f 87Sr/86Sr isotope ratios after removing a long-term trend92. g Carbon isotope composition (δ13C) values from marine sedimentary carbonates81 with a 15% locally estimated scatterplot smoothing (LOESS) trend line, compiled from regionally variable sources for the Carboniferous and predominantly from early Permian datasets93, representing a first-order approximation of long-term isotopic trends. h Modeled carbon emission rates (black area) and cumulative carbon release (red curve) derived from LOSCAR (Long-term Ocean-Atmosphere-Sediment CArbon cycle Reservoir) simulation. i Frequency histograms of global detrital zircon age counts22. j Extinction intensity across the late Paleozoic92. Phases I, II, and III denote tectonic phases defined in this study, corresponding to ~360–330 Ma, ~330–280 Ma, and ~280–250 Ma, respectively.

Full size image

Subduction zone and MOR lengths, together with surface area production rates (km2 Myr−1), provide direct constraints on lithospheric recycling, crustal accretion, and the reorganization of plate boundaries13. Peaks in global detrital zircon age densities further indicate pulses of continental arc magmatism and crustal addition22,28,29, while troughs reflect tectonic slowdown and reduced magmatic fluxes. 87Sr/86Sr values provide geochemical insight into the balance between continental weathering and MOR-derived mantle inputs. High 87Sr/86Sr ratios indicate increased contributions from continental sources, while low ratios are associated with enhanced MOR activity, offering critical insights into the global geochemical cycle and its linkage to tectonic dynamics30,31. Simultaneously, LOSCAR model results capture changes in carbon release magnitude and isotopic composition, offering an integrated perspective on how tectonic outgassing shaped atmospheric pCO2 levels and carbon isotope trends32,33,34 (Fig. 1a–d). Based on consistent cross-validation among these datasets, three tectonic phases can be recognized during the late Paleozoic, providing a framework for understanding the episodic nature of tectonic processes and their impact on magmatism and geochemical cycles. The intervals are defined as follows: 360–330 Ma (Phase I) and 280–250 Ma (Phase III) represent periods of elevated tectonic activity, while 330–280 Ma (Phase II) corresponds to a phase of reduced tectonic activity (Fig. 1). These phases are defined by relative changes in tectonic intensity rather than discrete geochronological boundaries, reflecting long-term trends in Earth system behavior.

Phase I (360–330 Ma) is marked by widespread subduction and MOR expansion, with detrital zircon frequencies reaching high values, indicating widespread continental magmatism. 87Sr/86Sr values trend lower, reflecting greater mantle input from seafloor spreading. In the LOSCAR simulations, this phase is represented by relatively elevated carbon input rates (averaging ~0.041 Pg C yr−1; Fig. 1h; Fig. S1), consistent with enhanced volcanic outgassing and active crustal recycling. Phase II (330–280 Ma) is characterized by a pronounced decline in subduction zone and MOR lengths, reduced crustal production rates, and troughs in zircon age distributions (Fig. 1). This interval shows elevated 87Sr/86Sr ratios, suggesting dominance of continental inputs and diminished MOR influence. In the carbon cycle simulations, this interval is represented by significantly lower carbon input rates (averaging <0.02 Pg C yr−1; Fig. 1h; Fig. S1), resulting in relatively stable or declining pCO2 concentrations. These factors collectively support an interpretation of this interval as a time of tectonic slowdown, with reduced volcanic degassing. This period was accompanied by the expansion and stabilization of large polar ice sheets, particularly in the Southern Hemisphere35. Phase III (280–250 Ma) marks a renewed phase of elevated tectonic activity. Subduction zone lengths and surface production increase, accompanied by resurgent magmatism recorded in detrital zircon peaks (Fig. 1). Concurrent decreases in 87Sr/86Sr ratios suggest intensified MOR input, consistent with enhanced mantle-derived strontium fluxes. Carbon input rates rise significantly during this phase in LOSCAR simulations (averaging ~0.063 Pg C yr−1; Fig. 1h; Fig. S1), reflecting a shift toward intensified mantle outgassing and large-scale magmatism. This resurgence likely contributed to the retreat of polar ice sheets and more pronounced global sea-level fluctuations36,37. While the precise transitions between phases are inherently gradual and somewhat uncertain, our classification emphasizes the existence of distinct phases defined by robust temporal trends rather than rigid geochronological cutoffs. We acknowledge the simplification involved in applying discrete phase boundaries, yet this framework is effective for identifying first-order relationships among tectonic background, carbon emissions, and climate sensitivity across the late Paleozoic Earth system.

Instability and obscuration of astronomical signals

Sea-level fluctuations offer a sensitive proxy that integrates multiple climate and geological processes operating across various timescales12,38,39 (Supplementary Information). On tectonic timescales, processes such as subduction and mid-ocean ridge expansion significantly impact global sea level by redistributing water between Earth’s interior and surface reservoirs12. At shorter, climatic timescales, sea-level variations primarily reflect the cyclic growth and retreat of ice sheets, thermal expansion of seawater, and changes in continental water storage, all of which are closely modulated by astronomical forcing through variations in solar insolation40,41,42. Therefore, orbital signals preserved in sea-level records represent a valuable means to investigate how tectonic activity might alter the expression and detectability of astronomical climate forcings. Clarifying this interaction is essential, as misinterpreting tectonically influenced sea-level changes solely as orbital signals risks distorting our understanding of past climate behavior and the underlying mechanisms driving global climate variability.

To assess how tectonic background influences the expression of astronomical forcing in sedimentary archives, short-period sea-level cycles across the late Paleozoic were analyzed using multiple lines of evidence. Statistically significant differences in sea-level cycle durations are observed when comparing intervals characterized by reduced tectonic activity (Phase II; ~330–280 Ma) with those exhibiting elevated tectonic activity (Phases I and III; ~360–330 and ~280–250 Ma) (Fig. 2; Fig. S3). To assess whether the observed cyclicities reflect periodicities consistent with astronomical control, circular spectral analysis (CSA) was performed on both sea-level cycle and sequence duration datasets. These results reveal that cyclicities within the range of 0.5–2.0 Myr are statistically significant during intervals of reduced tectonic activity, whereas signal coherence diminishes under more tectonically active conditions (Fig. 2; Fig. S3). Besides, during intervals of reduced tectonic activity (Phase II; ~330–280 Ma), sea-level cycles were consistently shorter and more tightly clustered (mean = 1.44 ± 0.61 Myr, n = 38). In contrast, periods of elevated tectonic activity (Phases I and III; ~360–330 Ma and ~280–250 Ma) were associated with longer and more dispersed cycle durations (mean = 1.85 ± 1.00 Myr, n = 31) (Fig. 2a). These differences are statistically robust, supported by Welch’s t-test (t = −4.44, p = 0.0001), F-test (F = 6.85, p < 0.0001), and one-way ANOVA (F = 23.03, p < 0.0001) (Supplementary Data 3). A similar pattern is also evident in the sequence stratigraphic records (Fig. S3; Supplementary Data 3).

Fig. 2: Sea-level cycle frequency and duration during periods of reduced and elevated tectonic activity.
figure 2

a Kernel density estimates of sea-level cycle durations (Myr) under two tectonic regimes: reduced tectonic activity (blue) and elevated tectonic activity (pink), corresponding to Phases II and Phases I + III, respectively. Red dots and horizontal bars denote the mean ± 1 standard deviation for each group. Inset rose diagram illustrates the distribution and concentration of cycle durations in circular space. b Circular spectral analysis (CSA) results of sea-level cycles during periods of reduced tectonic activity. Power spectrum (gold line) is plotted alongside Monte Carlo-derived confidence levels. c CSA results for sea-level cycles during elevated tectonic activity.

Full size image

Several plausible mechanisms may account for the observed differences in sea-level cycle durations and variabilities. Most notably, during tectonically reduced intervals, sedimentary basins likely experienced more stable accommodation space, lower sedimentation variability, and reduced environmental noise, facilitating clearer astronomical pacing. In contrast, elevated tectonic activity may have disrupted signal fidelity through enhanced basin subsidence, variable sediment supply, and localized deformation43 (Figs. 2, S3). These alterations might modify the boundary conditions of the climate system, thereby reducing its sensitivity to orbital forcing and diminishing the fidelity of preserved orbital signals19. Second, variations in long-term orbital modulation cycles or chaotic behavior within the solar system could potentially introduce uncertainties into sedimentary records44, although clear evidence supporting this hypothesis remains limited during the late Paleozoic. Lastly, biases and chronological uncertainties inherent in sea-level reconstructions might introduce variability; however, sensitivity analyses and multi-region cross-validation largely mitigate this possibility (Fig. S2). Although multiple mechanisms warrant consideration, the impact of varying tectonic intensity on the stability of orbital signals remains one of the most strongly supported explanations.

Late Paleozoic climate state variability

Climate variability, defined as the degree of fluctuation in climate variables such as temperature and precipitation, profoundly influences biological habitats, resource availability, and ecosystem stability45,46. To assess the relative contributions of continental configuration, orbital parameters, and atmospheric CO2 concentrations to climate state variability, targeted sensitivity experiments were conducted using CESM for three representative time slices: 340 Ma, 300 Ma, and 260 Ma (Fig. 3). These intervals span the entire duration of our study and capture contrasting tectonic and climatic regimes. The results indicate that variations in paleogeographic boundary conditions alone account for only minor differences in the spatial and temporal structure of climate variability (Fig. 3). In contrast, simulations with elevated atmospheric CO2 levels yield substantially greater variability in GMST and GMP (Fig. 4). These findings are consistent with the well-established principle that warmer climate systems exhibit heightened sensitivity to both internal variability and external forcings47,48. Therefore, the pronounced increase in variability during Phases I and III was likely primarily driven by elevated atmospheric CO2 concentrations (Figs. 4, S5–S8). Elevated atmospheric CO2 levels during Phases I and III likely arose predominantly from intensified tectonic degassing associated with enhanced subduction and volcanic activity (Fig. 1). Additionally, secondary mechanisms such as increased silicate weathering49,50 (which removes CO2 through chemical reactions) and reduced efficiency of organic carbon burial (which limits long-term carbon storage; Fig. 6) may further modulate atmospheric CO2 concentrations. However, a notable exception arises at 250 Ma, where elevated CO2 levels are paradoxically associated with a slight decrease in GMST and GMP (Fig. 4). A similar pattern has been observed in other modeling studies of the earliest Mesozoic, where reduced climate variability is reported around 250 Ma despite high atmospheric CO2 levels51. We hypothesize that this anomaly reflects the unique paleogeographic context of that time. Unlike earlier periods, the paleogeographic reconstruction for 250 Ma reveals nearly complete assembly of Pangaea and the closure of low-latitude seaways in the Tethys domain. These changes likely disrupted equatorial heat and moisture transport pathways52,53, thereby dampening the expected amplification of climate variability under high-CO2 conditions.

Fig. 3: Sensitivity of GMST variability to orbital forcing and atmospheric pCO2 at selected late Paleozoic time slices.
figure 3

Box plots show monthly global mean surface temperature (GMST, °C) simulated under different combinations of orbital configurations and atmospheric CO2 concentrations for 260 Ma, 300 Ma, and 340 Ma. At 260 Ma and 340 Ma, simulations were conducted under maximum orbital eccentricity and obliquity (Emax & Omax; eccentricity = 0.066, obliquity = 24.5°) with atmospheric pCO2 fixed at 400 ppm. At 300 Ma, four scenarios were tested: (1) Emax & Omax at 400 ppm CO2 (light blue), (2) Emax & Omax at 800 ppm CO2 (khaki), (3) Emin & Omin at 800 ppm CO2 (khaki; eccentricity = 0, obliquity = 22.5°), and (4) Emin & Omin at 400 ppm CO2 (light blue). Boxes represent the interquartile range (25th–75th percentile), red lines denote the mean GMST, and whiskers span the full range from minimum to maximum monthly values. Annotated arrows indicate the spread of monthly GMST variability under each scenario.

Full size image
Fig. 4: Late Paleozoic climate variability in temperature and precipitation (360–250 Ma).
figure 4

The temporal evolution of global mean surface temperature (GMST, °C) and global mean precipitation (GMP, mm/month) variability during the late Paleozoic (360–250 Ma) is derived from a suite of fully coupled Community Earth System Model (CESM) simulations conducted at 10-million-year intervals across 12 time slices. The red shaded area with box plots represents monthly variability in global mean surface temperature, while the blue shaded area with box plots corresponds to monthly variability in global mean precipitation. The dashed lines denote the mean values for each period, and the boxes indicate the interquartile range (25–75%). Six additional inset maps show zonal surface temperature differences and zonal precipitation differences for the time points 340 Ma, 290 Ma, and 250 Ma. CESM-simulated SST (sea surface temperature) averaged over paleolatitudes of ±5° and ±30° define the upper and lower bounds of the shaded band (dark blue), representing tropical SST variability across time. Conodont-based SST reconstructions (black crosses) are compiled from equatorial regions corresponding to paleolatitudes within the tropical belt, and the probability density distribution is visualized as a light blue background. The modeled SST consistently align with the empirical range, supporting the validity of CESM temperature outputs94. At the base, atmospheric pCO2 estimates and LOWESS trendline as in Fig. 1b.

Full size image

Atmospheric CO2 appears to play a central role in amplifying the climate system’s sensitivity to astronomical forcing. As demonstrated by the sensitivity experiments (Fig. 3), elevated pCO2 levels markedly increase GMST variability under a range of orbital configurations. Spatial patterns at 300 Ma (Fig. 5) further show that the differences in surface temperature and precipitation between high and low eccentricity–obliquity scenarios are significantly enhanced under elevated pCO2. These results indicate that elevated atmospheric CO2 intensifies radiative forcing and concurrently enhances the climate system’s sensitivity to insolation variability driven by orbital parameters. This enhanced sensitivity may be mediated by threshold behaviors and nonlinear feedbacks54,55, including abrupt glacial–interglacial transitions driven by ice volume sensitivity and ocean circulation reorganizations that modulate nutrient supply and biological productivity. Therefore, the observed climate instability during Phases I and III may mainly reflect a coupled response to rising atmospheric CO2 concentrations and orbital-scale insolation variability.

Fig. 5: Spatial differences in temperature and precipitation driven by orbital forcing under varying atmospheric pCO2 levels at 300 Ma.
figure 5

Maps illustrate the climatic response to different orbital configurations (Emax & Omax versus Emin & Omin) at 300 Ma, evaluated under atmospheric CO2 concentrations of 400 ppm and 800 ppm. a, b Annual mean surface temperature differences (°C) between Emax & Omax and Emin & Omin configurations. c, d Monthly mean precipitation differences (mm/month) under the same configurations.

Full size image

Organic carbon enrichment during icehouse period

Our synthesis of late Paleozoic organic-rich shale and coal deposits (Fig. 6) reveals a pronounced increase in deposition frequency during Phase II (~330–280 Ma), an interval characterized by relatively reduced tectonic activity compared to the preceding and following stages of intensified tectonism. Both volcanic activity and astronomical forcing are commonly identified as two crucial factors that influence organic matter accumulation56,57. Volcanic activity influences organic carbon dynamics through complex and bidirectional mechanisms. Moderate volcanism enhances nutrient availability by promoting silicate weathering and supplying trace elements such as phosphorus and iron, which together stimulate marine primary productivity50,58. In contrast, excessive volcanic episodes can perturb the climate system across multiple timescales. While short-lived eruptions may trigger transient cooling through stratospheric sulfate aerosol formation, particularly in the context of large igneous province volcanism, prolonged CO2 outgassing leads to sustained warming and environmental stress59,60. Over multimillion-year timescales, volcanism thus serves as a major source of greenhouse gases, reducing the preservation potential of organic matter through climatic destabilization61. Given that multiple geological indicators and carbon cycle model results suggest a substantial decline in volcanic activity during Phase II compared to Phases I and III (Figs. 1, S1). Such subdued volcanic conditions may have acted to enhance nutrient availability and primary productivity without triggering climate instability, thereby contributing positively to the observed organic matter enrichment. Astronomical forcing, in contrast, modulates climate variability by regulating seasonal and latitudinal patterns of solar insolation, thereby influencing temperature, precipitation, and hydrological cycling18,62. While the fundamental frequencies of orbital parameters are set by planetary mechanics and evolve on predictable timescales17,23, their climatic expression is strongly modulated by boundary conditions, including atmospheric CO2 levels, ice volume, and tectonic configurations (Figs. 3–5). Consequently, the climatic and depositional imprint of orbital cycles can vary substantially across geologic intervals. The observed increase in organic-rich sedimentation during Phase II, coupled with evidence of more coherent astronomical cyclicity during this interval (Fig. 2), suggests that reduced tectonic disturbance may have facilitated the clear imprint of astronomical forcing in sedimentary systems.

Fig. 6: Spatial and temporal distribution of organic-rich shales and coals during the late Paleozoic.
figure 6

a The distribution of organic-rich shales and coals across different paleolatitudes and time periods during the late Paleozoic. Each point represents a deposit from a marine, marginal, or lacustrine environment, with the color-coded dots corresponding to different depositional environments: blue for marine, green for MNT (Marine-nonmarine transition zone), and orange for lacustrine. Horizontal bars indicate age uncertainty associated with each deposit. The top histogram shows the frequency of organic-rich sediment deposition over time, while the side histogram shows their distribution by paleolatitude. The denser accumulation of deposits in tropical to subtropical regions (0° to 40° paleolatitude) is evident, especially during tectonically relatively stable periods such as Phase II. b The spatial distribution of organic-rich shales and coals during different tectonic phases. The majority of organic-rich deposits during Phase II are located in tropical and subtropical latitudes, with a relatively high density compared to the more broadly distributed records in Phases I and III (See Fig. S9 for details).

Full size image

We propose that comparatively reduced tectonic activity, in combination with low climate variability, played a key role in promoting organic matter accumulation during the late Paleozoic. Unlike the more geodynamically active intervals, Phase II (~330–280 Ma) coincided with an icehouse regime characterized by relatively low atmospheric CO2 levels, reduced climate variability in model simulations, and a more coherent expression of astronomical forcing (Figs. 3, 4). The cyclic patterns in sea-level variability and depositional architecture during this phase are commonly interpreted to reflect orbitally paced fluctuations in continental glacier volume and continental water storage42,63. Under reduced tectonic forcing, diminished crustal disturbance and enhanced accommodation continuity likely facilitated the preservation of such sedimentary rhythms, thereby improving the fidelity of astronomical signal expression in stratigraphic archives64,65,66 (Fig. 7a). This climatic backdrop created favorable conditions for biological productivity and rhythmic organic matter deposition, which facilitated the widespread development of organic-rich shales and coal beds in several regions, including North America, Europe, South China, and North China, sustaining an organic-rich icehouse period (Fig. 6b). Additionally, the spatial distribution of organic-rich shales and coals during this phase was predominantly concentrated between paleo-latitudes of 0° to 40°, although several occurrences are also observed between 40°S and 60°S (Fig. 6b). This pattern suggests that warm and humid equatorial to mid-latitude conditions facilitated both high productivity and effective preservation under relatively stable depositional settings. Within this framework, astronomical forcing likely paced climate fluctuations, further reinforcing the cyclic nature of organic carbon burial54,67. Collectively, the coupling of low-frequency climatic variability with enhanced orbital pacing provided optimal conditions for sustained organic matter accumulation during this critical interval.

Fig. 7: Schematic representation of tectonic and climatic influences on organic carbon burial.
figure 7

a During intervals characterized by lower tectonic intensity, relatively stable depositional settings and modest volcanic input may enhance the expression of orbital signals in sedimentary archives. These conditions support high biological productivity and promote the rhythmic accumulation and preservation of organic-rich sediments. b In contrast, intervals with elevated tectonic intensity are associated with increased subduction, mid-ocean ridge activity, and continental volcanism. These factors may amplify climate variability and disturb sedimentary environments through enhanced topographic reorganization, fluctuating nutrient fluxes, and altered hydrological regimes. These disruptions reduce biological productivity and the capacity for organic matter burial, leading to less organic-rich sediment accumulation.

Full size image

Elevated climate variability associated with periods of intensified tectonic forcing (Fig. 4), involving frequent fluctuations in temperature and precipitation, likely exerted substantial ecological stress through multiple interconnected pathways. Pronounced temperature fluctuations could periodically shorten or disrupt plant growing seasons, suppressing net primary productivity and consequently reducing the efficiency of terrestrial carbon sequestration capacity68,69. Similarly, greater precipitation variability likely amplified hydrological extremes (e.g., floods and droughts) during specific intervals. These extremes could further disturb sedimentary environments via abrupt sediment transport or erosion and impose long-term ecological stress, including nutrient loss and reduced resilience of biological communities to environmental fluctuations70. These environmental stressors did not universally suppress ecosystem productivity, as shown by flourishing ecosystems during certain warm intervals of the Cretaceous. In contrast, the late Paleozoic, characterized by extensive glaciation and large ice-volume variability1, ecosystems appear to have been more vulnerable to rapid environmental shifts71.

Further, intensified tectonic processes such as major volcanism, widespread mid-ocean ridge expansion, and continental orogeny during active intervals significantly modified Earth’s surface topography and ocean circulation patterns (Fig. 7b). These large-scale perturbations likely increased the variability and instability of regional climate systems, particularly through enhanced fluctuations in temperature and precipitation (Fig. 4). Such conditions may have impaired the recording and preservation of astronomical signals in sedimentary archives, as demonstrated by the broader and more irregular distribution of sea-level cycle durations during periods of elevated tectonic activity (Fig. 2; Fig. S3). A reduced number and fragmented distribution of organic-rich shales and coals observed during these intervals (Fig. 6) suggest significant disruptions to sedimentary continuity, consistent with increased depositional environmental instability caused by tectonic disturbances. These lines of evidence indicate that tectonic-driven environmental instability likely posed substantial challenges to ecosystems and sedimentary processes, ultimately limiting organic matter accumulation.

These results reveal the pivotal influence of tectonic conditions on how the climate system responded to astronomical forcing during the late Paleozoic. Intervals of reduced tectonic activity appear to have provided a relatively stable climatic background that facilitated coherent orbital pacing, enhanced biological productivity, and promoted widespread burial of organic matter in the form of shales and coals. In contrast, during intervals of elevated tectonic activity, the Earth system likely responded to astronomical forcing in more variable and nonlinear ways, amplifying climatic oscillations and disrupting depositional continuity. Notably, this variability does not imply instability in the orbital parameters themselves; instead, it reflects the dynamic response of the Earth system to external forcing under varying tectonic regimes (Fig. 7). While this study focuses on physical drivers of climate and carbon cycle interactions, the evolution of terrestrial vegetation likely also influenced organic carbon burial1,72. Changes in plant diversity and productivity, such as the expansion of seed plants and the collapse of tropical rainforests, may have affected carbon sequestration by altering biomass, soil formation, and sediment transport73,74. Although not included in the present modeling framework, such vegetation feedbacks are relevant to the broader Earth system and merit integration in future coupled simulations. Together with tectonic forcing, these biotic processes likely shaped how the climate system responded to astronomical variations. Recognizing this interaction provides a framework for interpreting carbon burial patterns in deep-time archives and offers useful analogs for other icehouse intervals.

Methods

Climate model experiments

Late Paleozoic climate dynamics (360–250 Ma) were investigated using the Community Earth System Model (CESM1.2.2), a fully coupled Earth system model incorporating atmosphere, ocean, land, and sea ice components75. The atmospheric and land components use the Community Atmosphere Model version 4 (CAM4) and Community Land Model version 4 (CLM4), respectively76,77, both configured with a horizontal resolution of 3.75° × 3.75° (T31 grid). The ocean (POP2) and sea ice (CICE4) components operate on a nominal 3° horizontal grid (g37), with 60 and 26 vertical layers, respectively78,79. Twelve simulations were conducted, each representing a 10-million-year time slice spanning from 360 Ma to 250 Ma (Fig. S4; Supplementary Data 4). Each run was initialized using reconstructed paleogeography consistent with the corresponding time interval. All other greenhouse gases, orbital parameters, and aerosol properties were held constant. Orbital parameters were fixed to a maximum eccentricity scenario (eccentricity = 0.066, obliquity = 24.5°, longitude of perihelion = 90°) to ensure consistent astronomical forcing across time slices. Each simulation was integrated for a minimum of 2000 years, and the final 100 years were used for analysis only after equilibrium was reached, defined by a global net top-of-atmosphere (TOA) radiative flux below 0.1 W m−2 (Fig. S4)80. These simulations reveal broadly consistent climate patterns across the Carboniferous and Permian. Surface temperatures and precipitation are highest near the equator and in tropical regions, whereas high-latitude zones remain cold and relatively dry (Fig. S5, S6).

In addition to the 12 baseline simulations, a series of targeted sensitivity experiments was conducted to examine the combined effects of orbital forcing and atmospheric CO2 concentrations (Fig. S4; Supplementary Data 4). At 260 Ma and 340 Ma, two simulations were performed under maximum eccentricity and obliquity conditions (eccentricity = 0.066; obliquity = 24.5°), with atmospheric pCO2 fixed at 400 ppm. At 300 Ma, four simulations tested the combined effects of orbital geometry and CO2 levels: (1) Emax & Omax at 400 ppm pCO2, (2) Emax & Omax at 800 ppm pCO2, (3) Emin & Omin at 800 ppm pCO2 (eccentricity = 0; obliquity = 22.5°), and (4) Emin & Omin at 400 ppm pCO2. These experiments provide critical insights into how orbital variability and CO2 radiative forcing jointly influence GMST variability and precipitation responses at orbital time scales (Figs. 3, 5).

LOSCAR carbon cycle model

Long-term carbon cycle dynamics during the Late Paleozoic were simulated using the LOSCAR (Long-term Ocean-Atmosphere-Sediment CArbon cycle Reservoir) model. LOSCAR is a box-model framework designed to capture the partitioning of carbon among the atmosphere, ocean, and sediments over long timescales, while explicitly tracking δ13C evolution in the surface ocean and atmospheric pCO2 levels32,33,34. The model consists of coupled modules representing ocean circulation, atmospheric exchange, biological productivity, weathering, and sedimentation32. In this simulation, LOSCAR includes 10 ocean boxes, subdivided into surface, intermediate, and deep layers for three major paleo-ocean domains and a high-latitude reservoir. All surface boxes are coupled to a single atmospheric box, enabling simulation of air-sea carbon exchange.

In the original LOSCAR design, emission forcing files contain only two columns: time and the carbon emission rate (Pg C yr−1). This format requires segmenting longer time-series runs whenever the carbon isotopic composition (δ13C) of emissions needs to vary through time. For studies involving geologically realistic, continuous changes in both emission rates and isotopic composition, particularly relevant to the late Paleozoic, this rigid structure constrains model flexibility and reduces computational efficiency. To improve input flexibility, we introduced a structural modification to the model input format, enabling the revised LOSCAR version to accept a three-column emission file specifying: (1) time (yr), (2) carbon emission flux (Pg C yr−1), and (3) δ13C of emitted carbon (‰). The model now directly integrates variable isotopic trajectories, eliminating the need to manually segment emission events into artificial steps with constant δ13C, which was previously necessary for long-duration, high-resolution simulations. The revised structure allows the model to continuously ingest time-resolved changes in both flux and isotopic composition from a single control file, improving simulation fidelity under geologically complex boundary conditions.

To adapt the LOSCAR model to the late Paleozoic interval, key initial parameters were revised to align with boundary conditions constrained by global geochemical and climate reconstructions (Supplementary Data 1). The initial global surface ocean dissolved inorganic carbon (DIC) δ13C was adjusted to 3.5‰, approximating observed late Paleozoic carbonate records81. The emitted arc volcanic CO2 is assumed to have δ13C=−3‰82. The modeled δ13C signal was derived by calculating the area-weighted mean δ13C across all surface ocean boxes, thereby integrating spatial heterogeneity while retaining consistency with marine carbonate proxy data. The final steady-state atmospheric CO2 concentration was set to 750 ppm to match geological CO2 proxy constraints and ensure mass balance in the silicate weathering–degassing cycle. Climate sensitivity was activated by enabling temperature feedbacks (TSNS = 1), and the climate response to CO2 doubling was set at 4.6 °C. All other physical and geochemical parameters (e.g., deep ocean temperature, biological pump efficiency, shelf-basin rain ratio) were maintained at either CESM-derived or default LOSCAR values, unless otherwise noted in Supplementary Data 1. The prescribed δ13C trajectory used in the emission input file was directly extracted from the global carbonate compilation presented in GTS202081. While this dataset integrates data from diverse depositional settings, including shallow-water carbonate platforms with potential subaerial exposure, it remains one of the few continuous δ13C records spanning the entire late Paleozoic interval. Despite possible biases during specific intervals such as the Carboniferous–Permian boundary, its temporal continuity and broad global scope make it suitable for long-term carbon cycle modeling.

Model execution follows a two-stage approach. First, a spin-up simulation under zero-emission conditions was performed for 2 Myr to ensure steady state (global net flux <10−3 Pg C yr−1). The resulting system state (LPIA_ini.dat) served as the baseline for all subsequent emission scenario simulations. Simulations were initiated using control files (e.g., whole.inp) that specified the desired time-varying emission fluxes and isotopic compositions through the modified three-column emission file. The results of the model runs are publicly available on Zenodo83. The objective of this modeling framework is to investigate relative trends in the Earth’s carbon cycle under tectonically- and orbitally-modulated forcing, rather than reproducing exact absolute values of each geochemical proxy. As such, minor discrepancies in ocean salinity, temperature distribution, and box boundary interactions, all of which are held constant in the model, are not expected to significantly influence the long-term trajectories of carbon isotope or atmospheric CO2 evolution. This approach enables robust, comparative analysis of carbon perturbation dynamics across varying tectonic and climatic states during the late Paleozoic.

Data compilation

The data used in this study were compiled from a comprehensive review of existing literature and stratigraphic records to inform the late Paleozoic climate state, organic matter burial patterns, and tectonic settings. The primary dataset includes organic-rich shale, coal, and lacustrine deposits from various basins around the world, spanning critical time intervals from the Early Carboniferous to the late Permian (360–250 Ma). This dataset is detailed in Supplementary Data 5, which categorizes the formations according to their paleolatitudinal positions, geological settings, and lithologies. These data provide key insights into how organic matter burial was distributed geographically and temporally during different phases of tectonic stability and activity. In this study, paleolatitudinal positions were derived from established paleogeographic reconstructions, and refined using the GPlates platform for paleolatitude recovery13,84.

Specifically, the formations and locations presented in the Supplementary Data 5 were selected based on the presence of organic-rich shales, coals, and lacustrine sediments across marine, transitional, and terrestrial environments. Data sources were compiled from peer-reviewed articles, including stratigraphic studies, paleogeographic reconstructions, and lithological analyses. These data informed the reconstruction of organic matter burial patterns, offering critical insights into how organic carbon accumulation coincided with periods of tectonic stability (Phase II, 330–280 Ma) and tectonic activity (Phases I and III, 360–330 Ma and 280–250 Ma, respectively). To ensure the reliability of the sea-level dataset, strict selection criteria were applied, including biostratigraphic constraints and lithological validation. Only formations with well-constrained radiometric dating or robust biostratigraphic correlations were included. Each entry was cross-referenced with lithological descriptions to confirm the presence of organic-rich strata, particularly focusing on shale, coal, and lacustrine deposits known for significant organic matter content.

Sea-level analysis and periodicity detection

To evaluate the existence of periodic signals in late Paleozoic sea-level variations, we combined classical peak-interval estimation with circular spectral analysis (CSA) and statistical variance testing. Our primary sea-level dataset is based on the widely adopted global eustatic curve developed by Haq and Schutter (2008), which provides biostratigraphically constrained sea-level fluctuations throughout the Carboniferous and Permian85. Although interpretive in nature, this curve remains the most widely used reference for multimillion-year-scale eustatic variations during the Carboniferous and Permian86,87. Temporal resolution and robustness were enhanced by supplementing the global sea-level curve with high-resolution regional stratigraphic sequences from the Tethyan domain and Donets Basin (Figs. S2, S3).

Short-period sea-level cycles were identified by calculating the time intervals between successive peaks in the detrended sea-level curve, as well as intersections between long-term and short-term components (Fig. 1c). This approach was also applied to the regional sequence datasets. The uncertainty introduced by manual identification of peaks and crossing points is negligible compared to the inherent resolution of the stratigraphic records. To statistically test for the presence of cyclicity in these discrete event sequences, CSA was applied to detect underlying periodic structures in event-based time series without requiring amplitude modulation88. CSA has been successfully used in analyzing periodicities in mass extinction events, impact craters, and other episodic geologic phenomena89. In this study, CSA was implemented using the Acycle 2.8 software90, which enables event-based Rayleigh power spectral analysis with Monte Carlo-based significance testing. The analysis was independently performed for short-period sea-level events associated with each tectonic phase. Statistically significant periodicities were identified by comparing Rayleigh power spectra against null distributions generated through stochastic simulations. Importantly, the analysis does not rely on the absolute amplitude or strict synchronicity of sea-level fluctuations, which are susceptible to local tectonics, basin subsidence, or glacio-eustatic coupling. Instead, we focus on the statistical distribution of cycle durations, which are less sensitive to absolute age calibration and better reflect the imprint of astronomical forcing under different tectonic regimes.

In parallel with spectral analysis, sea-level cycle duration distributions were statistically compared across different tectonic regimes to evaluate the impact of tectonic conditions on astronomical signal expression. The data were categorized into intervals of reduced and elevated tectonic activity (Supplementary Data 2). Distributions were tested using Welch’s t-test, F-test, and one-way ANOVA (Supplementary Data 3), and visualized through kernel density estimation and violin plots (Fig. 2a). This multi-method approach enabled us to detect orbitally paced signals and quantitatively assess how tectonic activity modulated the expression and preservation of astronomical forcing in sedimentary sea-level records.