CO₂-forced Late Miocene cooling and ecosystem reorganizations in East Asia

Significance

A widespread Late Miocene reorganization of terrestrial environments between 7.5 and 5 My B.P. occurred within a global context of ocean sea surface temperature cooling (up to 6 °C) and climate change. Yet the timing of events on land and the overall driving mechanism remains poorly understood. We present a new land surface temperature record from the Chinese Loess Plateau in East Asia, which reveals that cooling and aridification occurred synchronously with ocean cooling, highlighting a global forcing mechanism. Our results in combination with general circulation simulations suggest that positive feedbacks between decreasing atmospheric CO₂, global cooling, and enhanced land aridification promoted a major reorganization of ecosystems in East Asia during the Late Miocene.

Abstract

In parallel with pronounced cooling in the oceans, vast areas of the continents experienced enhanced aridification and restructuring of vegetation and animal communities during the Late Miocene. Debate continues over whether pCO₂-induced global cooling was the primary driver of this climate and ecosystem upheaval on land. Here we present an 8 to 5 Ma land surface temperatures (LST) record from East Asia derived from paleosol carbonate clumped isotopes and integrated with climate model simulations. The LST cooled by ~7 °C between 7.5 and 5.7 Ma, followed by rapid warming across the Miocene–Pliocene transition (5.5 to 5 Ma). These changes occurred synchronously with variations in alkenone and Mg/Ca-based sea surface temperatures and with hydroclimate and ecosystem shifts in East Asia, highlighting a global climate forcing mechanism. Our modeling experiments additionally demonstrate that pCO₂-forced cooling would have altered moisture transfer and pathways and driven extensive aridification in East Asia. We, thus, conclude that the East Asian hydroclimate and ecosystem shift was primarily controlled by pCO₂-forced global cooling between 8 and 5 Ma.

The Late Miocene (11.6 to 5.3 Ma) represents a predominantly warmer-than-present interval that was characterized by elevated atmospheric CO₂ levels (pCO₂) and much reduced polar ice sheets (1). Between 8 and 5 Ma, the Earth’s climate experienced upheaval both in the ocean and on land and shifted from more equable, warmer conditions toward a near-modern climate with dynamic glaciations in the Northern Hemisphere (2, 3). This period, thus, offers the opportunity to assess the repercussions of climate fluctuations on terrestrial ecosystems and is highly relevant to better constrain scenarios of future global warming (1).

Currently, the dynamic processes and driving forces behind the Late Miocene climate shift are still vigorously debated (2–4). Alkenone and Mg/Ca-based sea surface temperature (SST) reconstructions reveal prominent cooling in both hemispheres from 7.8 to 5.8 Ma, i.e., Late Miocene cooling (LMC), followed by a rapid SST rebound in the latest Miocene to earliest Pliocene (5.5 to 5 Ma) (2, 3). Ocean cooling was concurrent with enhanced aridification (5–8), marked by expansion of global C₄ grasslands (plants using the C₄ photosynthetic pathway) (9–11), and restructuring of mammal communities (12) on land. While some previous studies inferred that a decline in pCO₂ gradually drove ocean cooling (2, 3) and terrestrial ecosystem changes (10, 11) during the Late Miocene, other studies invoked paleogeographic changes [e.g., ocean gateways closure (4, 8) and regional mountain uplift (13)] as the main triggering mechanisms. Disentangling these potential drivers is especially challenging in Asia where this extensive climate transition during the LMC may have been concurrent with regional uplift of the northeastern margin of the Tibetan Plateau (14) and/or rain shadow development behind the Central Asian Orogenic Belt (15). Moreover, the poorly constrained evolution of terrestrial temperatures in Asia over the Late Miocene (16, 17) prohibits assessment of the synchroneity of climatic events on land and in the ocean as well as evaluation of potential relationships to a common pCO₂ forcing.

Results

Here, we present a land surface temperature (LST) record for the period 8 to 5 Ma (i.e., Late Miocene to earliest Pliocene) from the northern Chinese Loess Plateau (CLP) in East Asia (Fig. 1A and SI Appendix, Figs. S1–S3). New paleosol carbonate nodule T(Δ₄₇) data (n = 10), derived from clumped isotope (Δ₄₇) analysis, were integrated with a recalibrated published Δ₄₇ data set (n = 12) from a previous study (18) to extend the LST record over the interval 8 to 5 Ma (see Methods and Dataset S1), in order to determine the magnitude and timing of LST variations and their relationship to global SSTs and East Asian hydroclimate and ecosystem changes. Overlapping measurements indicate that our new T(Δ₄₇) results are congruent with the recalibrated published data (18) (Dataset S1 and SI Appendix, Fig. S4). The T(Δ₄₇) and calculated soil water δ¹⁸O_sw in all samples range from 9 °C (± 2 °C) to 19 °C (± 3 °C) (1 SE) and from −10 to −8.5‰ (Vienna Standard Mean Ocean Water, VSMOW), respectively (Fig. 1B and SI Appendix, Fig. S5). A change point analysis was performed to identify the mean trend of T(Δ₄₇) evolution between 8 and 5 Ma (see Methods and SI Appendix, Fig. S5). The T(Δ₄₇) results indicate an average temperature of 17 °C (± 4 °C) in the later part of the Tortonian (8 to 7.3 Ma), which is ~6 °C warmer than the study area’s present mean annual soil temperature (MAST, ~10 °C) (Fig. 1B). Between 7.5 and 6.7 Ma, T(Δ₄₇) values dropped by ~7 °C from 17 °C (± 4 °C) to 10 °C (± 2 °C) and hovered around 9 to 11 °C, which is close to present MAST, until 5.7 Ma (SI Appendix, Fig. S5A). Temperature then rebounded to 14 °C (± 2 °C) across the Miocene–Pliocene transition (~5.5 to 5 Ma).

Fig. 1.

LST versus compiled SST, foraminiferal δ¹⁸O, δ¹³C and pCO₂ evolution between 8 and 5 Ma. (A) Location of the CLP (square) in East Asia. Colored circles indicate the locations of records used in the global comparison shown in Figs. 1 and 2. (B) Clumped isotope (Δ₄₇)-derived LST from the northern CLP, combined new data, and recalibrated published data from ref. (18), error bars = 1 SE. The dashed line marks the present MAST in the study area. (C) Standardized tropical (red circles in Fig. 1A), mid-latitude (30 to 50°N/S; orange circles), and high-latitude (>50°N; blue circles) SST composite records during the 8 to 5 interval (see Method in Supplemental note 2 and SI Appendix, Figs. S6 and S7). Source data are shown in Dataset S2. (D) Evolution of planktic foraminiferal δ¹⁸O at ODP Site 1146 in the South China Sea (3). The TG22-TG12 events indicate peak glacial stages (19). (E) Gradient (Δδ¹³C) between benthic and planktic foraminiferal δ¹³C in ODP Site 1146 (3). (F) δ¹¹B (20, 21) and alkenone-based (22, 23) pCO₂ estimations. Smooth curve fitted in Fig. 1C using locally weighted least squared error (LOWESS) method. The blue bar indicates the development of transient Northern Hemisphere ice sheets (19, 24).

Comparison with Existing Records and Discussion.

We interpret the T(Δ₄₇) values to be primary and track secular MAST change in the northern CLP. The homogeneous dense micrite samples show no signs of fluid-mediated alteration (SI Appendix, Figs. S3 and S5) and were too shallowly buried (<260 m; 18) to undergo solid-state reordering due to post-depositional heating. The carbonate isotopic values should therefore reflect equilibrium with regional soil water δ¹⁸O_sw and surrounding soil temperature over timescales of centuries to millennia (25) during the season(s) in which soil carbonate accumulation occurred. Soil carbonate-bearing paleosol layers throughout the sampled “red clay” deposits are composed of fine-grained (clay and silt) matrix, which retains moisture better than coarse-grained soils and may delay soil drying and carbonate accumulation until fall, when soil temperature is close to MAST (26). We therefore interpret the T(Δ₄₇) values to represent MAST, which is comparable to terrestrial mean annual air temperature (MAT) for samples collected at depths >20 cm (26). The Late Miocene T(Δ₄₇) record (8 to 5 Ma; range 9 to 19 °C; mean 13 °C) from the northern CLP is consistent with coeval microbial branched glycerol dialkyl glycerol tetraether-derived MAT from the Mediterranean Basin [6 to 21 °C; mean 13.4 °C; (7)] and nearby Xining Basin [10 to 14 °C; (16)], and phytolith-derived MAT from the Weihe Basin in the southern CLP [11 to 15 °C; (17)], supporting this interpretation. Furthermore, we suggest that the detected T(Δ₄₇) drop since the late Tortonian is not due to a change in soil carbonate accumulation season (i.e., shift from warm season-biased to fall) or other bias. First, the uniformly fine paleosol grain size predicts no systematic change in soil drainage/evaporation or CO₂ diffusion coefficient that would alter carbonate accumulation timing. Extremely arid localities (i.e., mean annual precipitation <300 mm) with bare soils devoid of vegetation might allow for carbonate accumulation during the summer regardless of soil texture, or soil temperatures at the sampled depths that exceed MAT by 0 to 3 °C due to radiative heating (26), but phytoliths reconstruction suggests that the CLP was not extremely dry and completely bare soil during the Late Miocene (17). Second, in some studies it has been suggested that a change in soil carbonate accumulation season may be accompanied by estimated soil water δ¹⁸O_sw shift (27, 28). However, the reconstructed soil water δ¹⁸O_sw from which the paleosol carbonates grew exhibits a largely uniform range and is in broad agreement with present mean annual precipitation δ¹⁸O (SI Appendix, Fig. S5), providing no indication of a shift. Finally, existing mid-latitude terrestrial and marine reconstructions for the Tortonian stage and early Pliocene estimated a MAT warming of ~5 to 8 °C and ~3 °C, respectively, relative to the present (2, 7, 16). Thus, our late Tortonian and mean T(Δ₄₇) values, which are ~6 °C and 2 to 3 °C warmer than the study area’s present MAST, respectively, are within a reasonable range of warmer-than-present MAST and require no shift in carbonate seasonality (SI Appendix, Supplemental note 1). We therefore consider our T(Δ₄₇) record to reflect the regional MAST and track secular terrestrial climatic change over the 8 to 5 Ma interval [see SI Appendix, Supplemental note 1 for details of paleoclimatic implication of soil carbonate T(Δ₄₇) variations].

The evolution of LST from the northern CLP (~7 °C cooling since the late Tortonian and subsequent warming in the earliest Pliocene) is consistent with the trend and magnitude (up to 6 °C) of mid-high-latitude annual SST changes in both hemispheres [Fig. 1C and SI Appendix, Figs. S6 and S7; (2)] and with the trend of the planktic δ¹⁸O stack [Fig. 1D; (3)]. Synchronous cooling of similar magnitude (8 °C) is also recorded in the nearby Japan Sea Site U1425 (39.5°N, Fig. 1A), where the annual SST dropped from 24 °C to 16 °C between 7.9 and 6.6 Ma (29). During the period of maximum cooling (i.e., ~7 to 5.7 Ma), the LST was intermittently close to the present MAST (10 °C, Fig. 1B). This cooling pattern, which brought SSTs to approximately their near-modern values in the Messinian (~7 to 5.4 Ma), is clearly expressed in the mid-high-latitude Pacific Ocean (2). Moreover, this pronounced cooling interval in the terrestrial record (Fig. 1B) corresponds to the first emergence of transient Northern Hemisphere glaciations during a ~2-My time window [~7 to 5 Ma, (24)], culminating with multiple episodes of benthic and planktic δ¹⁸O maxima marking peak glacial stages [TG22-TG12 events between 6 and 5.4 Ma; Fig. 1D; (3, 19)]. Therefore, we hypothesize that a common driving mechanism (e.g., pCO₂ forcing) led to consistent temperature changes both on land and in the oceans during the LMC.

The long-term, high-magnitude LST cooling in the CLP since 7.5 Ma (i.e., LMC) was accompanied by drying and by ecosystem shifts on the vast Asian landmass (Fig. 2 A–C). During the LMC, soil humidity decreased in the North China Plain, the xerophytic herbs and deserts increased in Northern China, and the intensity of chemical weathering decreased in the South China (these records and references are shown in the SI Appendix, Figs. S8 and S9). In addition, precipitation also decreased in the Himalayan foreland and Arabian Peninsula, as shown by the sustained increase in hydrogen isotopic values of plant-wax n-alkanes (30). As a result of the cooling and drying, intensified dust emission in the Asian interior caused increased eolian dust fluxes into the North Pacific Ocean (31, 32) and the South China Sea (SCS) [Fig. 2B; (33)]. In Northwestern China, the source regions of the eolian dust expanded, and the dust accumulation rate and the proportions of the coarse-grained (>30μm) component and soluble salts also increased (see detail in SI Appendix, Fig. S8). This climatic and environmental stress resulted in profound shifts in ecosystem structures. The proportion of cold-arid adapted mollusk species in the CLP increased at 7.1 to 6.0 Ma [Fig. 2D; (34)]. In Southeast Asia, this marked climate cooling and drying caused a major change in biota, including the disappearance of hominin lineages (35) and a decrease in terrestrial mammalian ecological diversity [Fig. 2D; (12)].

Fig. 2.

Comparison of terrestrial temperature evolution with hydrological and ecological changes between 8 and 5 Ma. (A) Clumped isotope (Δ₄₇)-derived LST from the northern CLP, combined new data, and recalibrated published data from ref. 18. (B) Eolian dust fluxes recorded in ODP Sites 885 (31), 1208 (32) and 1146 (33). Note: the dust fluxes in ODP Site 885 may require further revision following Abell et al. (36). (C) Hydrological evolution of EASM, recorded in the SCS indicated by chemical weathering intensity (CIA and Rb/Sr, 29), δ¹⁸O_seawater (37), and alkenone C₃₇ concentration (38). Note: the records of chemical weathering intensity may also potentially be impacted by the onset of the Taiwan Orogeny at ~5 Ma (38). (D) Evolution of mollusk species in the CLP (34), mammal species in Pakistan of Southeast Asia (updated from ref. 12), and disappearance time of hominoid lineages in South China (rhomb, 35). (E) Range of tooth enamel δ¹³C from North American below 37°N (shading, 10); δ¹³C of leaf wax C₃₁ n-alkane in ODP Site 718 (30). (F) C₄ diet of ungulates from China (shading, 9); minimum C₄ grasses (%) estimation from CLP phytolith (17), and δ¹³C of C₃₁ n-alkane from ODP Site 722 (39). Smooth curve fitted using ten-point LOWESS method.

The East Asian Summer Monsoon (EASM) is responsible for delivering seasonal moisture from the western Pacific Ocean and Intertropical Convergence Zone (ITCZ) to the East Asian landmass (40). Thus, the monsoonal precipitation belt over the mainland is closely related to the location of the ITCZ and western flank of the western Pacific subtropical high (WPSH). Cooling of the Asian landmass would lead to a southward drift of the ITCZ (41) and a shrinkage of the WPSH (40), resulting in decreased tropical convection and intensified dry winter monsoon over southern Asia, as reflected by a substantial increase in seawater δ¹⁸O after 7.5 Ma in the SCS [Fig. 2C; (3, 37)]. The documented cooling of the Asian landmass was associated with a long-term weakening of the EASM, as indicated by chemical weathering intensity indices [chemical index of alteration (CIA) and Rb/Sr] records at Ocean Drilling Program (ODP) Site 1146 in the SCS [Fig. 2C; (33, 42)]. Moreover, multiple geochemical proxies derived from the nearby ODP Site 1148 in the SCS show reduced chemical weathering intensity since 7.5 Ma due to decreased humidity from the weakened monsoon (43). In addition, the extremely low productivity and weak upwelling in the SCS, reflected by the lowest alkenone C₃₇ values at 6.5 to 5.6 Ma, also indicate weak summer monsoonal transport over this relatively colder climate interval [Fig. 2C; (38)].

A decrease in atmospheric pCO₂ in the later part of the Late Miocene is supported by several independent lines of evidence. Although uncertainties remain large, the latest pCO₂ reconstructions based on different methods (e.g., foraminiferal δ¹¹B and phytoplankton alkenone) indicate a decline in pCO₂ between 7.5 and 5.5 Ma (Fig. 1F;  20–23; see further details in SI Appendix, Supplemental note 3 and Fig. S10). A carbon-alkalinity-calcium box model used to assess the climate feedback of giant evaporite deposition on the carbon cycle during the Messinian salinity crisis indicated that evaporite deposition would result in >50 ppm pCO₂ decline, amplifying the cooling trend during the Messinian (7, 44). Furthermore, a drop in CO₂ level during the Messinian is still the most plausible scenario for the emergence of vital effects in coccoliths (45) and transient Northern Hemisphere glaciations after 7 Ma (2, 24). These observations suggest that pCO₂ forcing may have exerted a considerable impact and driven synchronous LST-SST variations. The vertical δ¹³C gradient between benthic and planktic foraminifers (denoted as Δδ¹³C) can also be used as an approximate indicator of pCO₂ change (3), as lower pCO₂ levels lead to elevated δ¹³C values for the mixed layer and a relatively larger Δδ¹³C (3; SI Appendix, Supplemental note 3). The synchroneity of the LST response with the Δδ¹³C evolution at ODP Site 1146 in the SCS (Fig. 1E), thus, additionally supports a linkage between the global marine carbon cycle and terrestrial cooling.

In the context of the large-scale LST-SST cooling and increase in aridity, terrestrial C₄ grasses in East Asia (9, 17, 46), North America (10), the Mediterranean region (7), and Africa (11) show expansion during the LMC (Fig. 2 E and F). Phytolith data from the CLP in East Asia indicate that C₄ grasses (Panicoideae-Chloridoideae) exhibit increasing trends starting at ~7.2 Ma (17; Fig. 2F). Accelerated cooling itself is actually unfavorable for C₄ grass expansion (9), but this effect is offset by other competitive advantages of the C₄ photosynthetic pathway, which is more efficient in atmospheric CO₂ uptake and in water retention under lower pCO₂ levels (9). A similar large-scale photosynthesis adjustment was also identified in marine coccolithophorid algae at ~7.0 Ma, which has been interpreted as a threshold response to decreased pCO₂ levels (45). Taken together, we view the scenario of synchronous climate cooling and change in carbon acquisition strategies in both terrestrial and marine settings as providing strong support for the hypothesis (10) that the decline in pCO₂ during the Late Miocene favored global C₄ grassland expansion. In addition, the pCO₂-forced LST cooling would have increased aridity-related fire activity (SI Appendix, Fig. S8), further accelerating the regional preconditions for the expansion of C₄ grasses in East Asia.

Across the Miocene–Pliocene transition (between 5.5 and 5 Ma), the drying trend on land was reversed simultaneously in both hemispheres (47, 48). Over most of East Asia, the cold and dry climate progressively gave way to warm and humid condition, as corroborated by the LST warming in the northern CLP (Fig. 2A), dust fluxes in the North Pacific Ocean and SCS (31–33; Fig. 2B), and hydroclimate proxies in the East Asian monsoon region (Fig. 2C and SI Appendix, Fig. S8; also see compilation in ref. 48). The consistent warming between 5.5 and 5 Ma favored the expansion of warm-humid adapted mollusks in the CLP [Fig. 2D; (34)]. The pollen record additionally reveals the resumption of woodland (i.e., broadleaf trees) in Central and East China, which also suggests that a relatively warm and humid climate prevailed from the end of the Messinian to the earliest Pliocene (SI Appendix, Fig. S9). These synchronous trends in terrestrial temperature, hydroclimate, and ecosystem change lead us to speculate that pCO₂-induced global cooling exerted a major role in modulating moisture penetration onto the East Asian landmasses between 8 and 5 Ma.

To evaluate the hydroclimate response and feedback mechanisms involved with a decline in pCO₂ over East Asia, we conducted a series of idealized experiments that sample potential atmospheric pCO₂ of 400 ppm, 284 ppm, and 142 ppm, using the CESM model (version 1.2) (Fig. 3 and SI Appendix, Fig. S12). These pCO₂ ranges are inferred from compiled reconstruction results (Fig. 2F and SI Appendix, Fig. S10). The decline in pCO₂ results in an overall annual cooling of 2 to 4 °C in most regions of East Asia and amplified cooling (4 to 6 °C) in the mid-high latitudes, which leads to a relatively steep meridional temperature gradient (SI Appendix, Fig. S13). The sensitivity experiments with different magnitudes of pCO₂ decline show a decrease in land-sea thermal contrast between the Asian mainland and the equatorial western Pacific. In summer, the East Asian monsoonal circulation weakens across the region north of 40°N, as atmospheric pCO₂ declines from 400 to 284 ppm, and further weakens over East Asia at 142 ppm. The weakening of the East Asian monsoonal circulation leads to a significant decrease in moisture transport from the SCS and the West Pacific Ocean to the mid and high latitudes of East Asia, coupled to a decrease in summer precipitation (Fig. 3 and SI Appendix, Fig. S14). Although the monsoonal circulation is strengthened over South China as pCO₂ declines, the enhanced vapor divergence still contributes to the drying of this region. Overall, our modeling results indicate that the pCO₂-induced global cooling forces a weakened moisture transport (including both circulation and humidity) and vapor convergence, enhancing land aridity over East Asian modern monsoon regions.

Fig. 3.

Simulated changes in summer (June-July-August) precipitation and circulation over East Asia when pCO₂ is decreased from 400 ppm to 284 ppm and then to 142 ppm. (A and B) changes in summer precipitation (shaded, unit: mm/d) and wind fields (vectors, unit: m/s) at 850 hPa. (C and D) changes in summer vertically integrated moisture transport (vectors, units: kg m⁻¹ s⁻¹) and vapor divergence (shaded, units: 10⁻⁵ kg m⁻² s⁻¹).

We chose to conduct the modeling under preindustrial (PI) boundary conditions (e.g., as done in refs. 49 and 50) rather than under more realistic ones (i.e., late Neogene), because in this way, the effect of pCO₂ itself can be isolated and more clearly assessed. As all else is well vetted and held equal, this approach allows for a more confident estimation of the impact of the pCO₂ decline. Such an idealized simulation is reasonable, since key features of Asian topography reached near-modern conditions before the latest Miocene, in particular the Tibetan Plateau and Himalaya (14, 51–54), which provide strong orographic forcing for East Asian hydroclimate (55). Moreover, the East Asian hydrological response to pCO₂ forcing appears to follow a consistent pattern under late Neogene and near-modern boundary conditions (40, 41, 48, 56). For example, applying mid-Pliocene boundary conditions with pCO₂ forcing only, the simulated hydrological evolution shows an increase in summer net moisture transport over most of East Asia, when pCO₂ rises from 280 to 400 ppm (SI Appendix, Fig. S15A; 48). Simulation under Late Miocene boundary conditions also suggests wetter condition in the East Asian monsoon region in responses to CO₂-induced global warming (SI Appendix, Fig. S15B; 56). The similar hydrological responses under both our and previous simulations underscore the essential role of CO₂-induced global warming in forcing hydrological changes over the East Asian monsoon regions since the Late Miocene, through the decreased land-sea thermal contrast and diminished atmospheric water vapor content (Fig. 3). Therefore, our results imply that the East Asian monsoon regions may also experience increasingly wetter conditions, as the Earth continues to warm due to rising atmospheric pCO₂ levels.

Besides pCO₂ forcing, the shrinkage of the Tethys seaway (8) and reactivation of the Central Asian Orogenic Belt (15) during the Tortonian (~11 to 7 Ma) could also have influenced regional atmospheric moisture transport, leading to enhanced aridification in North Africa and in the Asian interior. However, the subsequent resurgence of temperature and humidity in both hemispheres (47, 48) is difficult to reconcile with this irreversible regional tectonic forcing. Although modeling studies demonstrated a connection between Tibetan Plateau uplift, inland aridification, vegetation dynamics, and increased CLP eolian accumulation over inner Asia (55), a large part of the plateau region was already elevated before the end of Miocene (14, 51–54). Even in the context of rejuvenated regional tectonic activities at the northern margin of the Tibetan Plateau, the high-resolution climate record from the Tarim Basin in Western China shows that regional moisture patterns were regulated by global change until the earliest Pliocene (~4.9 Ma) (57). More crucially, it seems difficult to directly link the synchronous LST-SST cooling and corresponding increased aridity in Eastern Asia to the topographic growth of the Northern Tibetan Plateau, given that such uplift would have induced an increase in winter precipitation and a transition to a more humid climate over Eastern Asia (55, 58). Thus, we consider pCO₂-forced cooling to have played a fundamental role in driving the Late Miocene East Asian climate shift. Paleogeographic changes in inner Asia may have amplified the atmospheric moisture effects, rather than initiate the climate shift in East Asia during the Late Miocene–earliest Pliocene (8 to 5 Ma).

Conclusion

In synchroneity with SST variations, the LST record from the northern CLP in East Asia reveals a sharp cooling between 7.5 and 5.7 Ma followed by a rebound across the Miocene–Pliocene transition (5.5 to 5 Ma). The coupled evolution of LST-SST and the associated hydrological and ecosystem changes in East Asia support a global forcing mechanism by atmospheric CO₂ change from the Late Miocene to earliest Pliocene (8 to 5 Ma). Our combined empirical and modeling results provide additional insights into projected climate development on Earth, suggesting that wetter conditions will develop over East Asian monsoon regions in an anthropogenically forced warmer future.

Materials and Methods

Geological Background and Materials.

The studied “red clay” sequences at the Baode (39°N, 111°E) and Shilou (37°N, 11°E) sites are situated in the northern CLP in East Asia. These sequences, which consist of alternating silty loess and paleosols layers, provide a virtually continuous sedimentary record since ~8 Ma (SI Appendix, Fig. S1). The age framework of the Baode and Shilou ‘red clay’ sequences (Late Miocene to Pliocene eolian deposits) is based on high-resolution magnetostratigraphy (SI Appendix, Fig. S2). The timescale for our samples was built by linearly interpolating between age and depth control points, which were obtained based on magnetostratigraphic boundaries, resulting in an age model uncertainty of <10 kyr (SI Appendix, Fig. S2). Carbonate nodule-bearing paleosol layers are relatively abundant throughout the eolian “red clay” deposits and thus archive a continuous history of climatic change. To reduce the potential effect of radiative heating on the LST results, we collected carbonate nodules (1 to 2 cm in diameter) in paleosol layers at depths >30 cm. Currently, the study area is characterized by a MAST of ~10 °C at 20 cm soil depth, with mean summer (June to September) soil temperature of ~22 to 24 °C and air temperature of ~20 °C, respectively (based on China Meteorological Data Service Centre record during 1951 to 2001, http://data.cma.cn). The modern average annual rainfall of 400 to 450 mm in this region is controlled by the EASM, with over 70% of precipitation occurring during the summer.

Cathodoluminescence (CL) Microscopy.

Paleosol carbonate nodules were made into polished thin sections. We evaluated the preservation of paleosol carbonate nodules with polarized light microscopy and CL. Cathodoluminescence microscopy petrography was performed using a CITL 8200 Mk3, Optical Cathodoluminescence System. Nodules exhibiting primary, dense, and homogeneous micrite with rare coarse siliciclastic grains/secondary veins were selected for clumped isotope analysis (SI Appendix, Fig. S3).

Clumped Isotope Analysis.

We measured the clumped isotopic composition of CO₂ (a measure of the abundance of ¹³C–¹⁸O bonds in carbonates relative to the abundance expected if the isotopologues were stochastically distributed; reported as Δ₄₇) derived from carbonate-derived CO₂ to determine the formation temperature of pedogenic carbonates nodules (59). For each sample, ~50 to 100 mg carbonate powder was drilled from the polished carbonate nodule surface to a depth shallower than 2 mm. Δ₄₇, δ¹⁸O, and δ¹³C were measured concurrently with three to seven replicates per sample at the University of Washington IsoLab, Seattle, USA, following ref. 60. For each replicate analysis, ~10 mg of carbonate powder was acidified in a common bath of 90 °C phosphoric acid for ~10 min, and the generated CO₂ was purified by passage through a series of cryogenic traps. Intra and interlaboratory carbonate standards C64, C2, Coral, and ETH 1-4 were analyzed to monitor long-term analytical error and to place carbonate δ¹³C and δ¹⁸O_c values on the VPDB (Vienna Pee Dee Belemnite) scale. Δ₄₇ values were calculated using IUPAC (International Union of Pure and Applied Chemistry) parameters to correct for ¹⁷O mass interference. The calibration of Anderson et al. (61) was used to generate T(Δ₄₇): Δ_{47(I-CDES90 °C)} = (0.0391 ± 0.0004)*(10⁶/T²) + (0.154 ± 0.004) [T in kelvin]. The subscript CDES means carbon dioxide equilibrium scale standardization scheme (62). Reported Δ₄₇ uncertainties include the error contribution from the CDES reference frame. External SE (1SE) in Δ₄₇ analyses is calculated as the long-term SD of carbonate standard analyses, or of sample replicates, whichever is larger, divided by the square root of the number of replicates for each sample. Complete Δ₄₇ data are reported in Dataset S1 and archived in the EarthChem repository. Soil water δ¹⁸O_sw values were calculated from the clumped temperatures T(Δ₄₇) and soil carbonate δ¹⁸O_c values (Dataset S1) and shown in SI Appendix, Fig. S5.

Updating Published T(Δ₄₇) Values.

Suarez et al. (18) also reported median T(Δ₄₇) values ranging from 16.5 to 20.1 °C for Late Miocene (~7.2 to 5 Ma) paleosol carbonates from Baode in the northern CLP. Their samples were digested at 90 °C, and an acid fraction factor (AFF) of 0.081‰ was applied. Their Δ₄₇ values were calculated using an older parameter set than the IUPAC values for ¹⁷O correction and reported relative to the old “Ghosh” reference frame scale (63) rather than the CDES scale and converted to temperature using the calibration of Ghosh et al. (Δ₄₇ = 0.0592*10⁶/T²-0.02 [T in kelvin]) (63). The Δ₄₇ values were also reported relative to the CDES [also known as “absolute reference frame” scale] (62), but these values were not used to calculate T(Δ₄₇) (see Supplementary Table DR3 of ref. 18).

At present, the “Ghosh” scale and calibration used by Suarez et al. (18) are no longer recommended. Replicate analysis, methods of standardization, and calibrations from recent studies significantly improved the agreement of the T-Δ₄₇ relationship (64). Here, the following steps are followed to update ∆₄₇ values and temperature estimates from Suarez et al. (18). We adopt Δ₄₇ values relative to the CDES scale and use the newly published AFF of +0.088‰ (64) instead of the original value of 0.081‰ (63) to convert 90 °C-reacted data to the CDES25 scale. The necessary raw data are lacking to reprocess the Δ₄₇ data using updated IUPAC ¹⁷O correction parameters; therefore, following Petersen et al. (64), we use the most up-to-date, CDES-referenced, lab-specific T-Δ₄₇ calibration based on data collected in the same laboratory and calculated in identical manner to the Suarez et al. (18) samples: Δ_47-RFAC = (0.0388 ± 0.0004)*(10⁶/ T²)+(0.262 ± 0.005), with an r² value of 0.99, T in kelvin [reported in SI Appendix, Table S5 of Petersen et al. (64)]. The subscript RFAC means absolute reference frame with acid corrected. As shown by Petersen et al. (64), T(Δ₄₇) values calculated in this way are comparable to the new T(Δ₄₇) values we report (SI Appendix, Fig. S4 and Dataset S1). The median value of the updated T(Δ₄₇) ranges from 9 to 14 °C (SI Appendix, Fig. S16), which is consistent with that of our newly added analyses in the overlapping portion of the record (from 10 to 17 °C; SI Appendix, Fig. S4). The reduction in final T(Δ₄₇) in all samples and consideration of fine-grained soil texture changes the interpretation of the soil carbonate formation seasonality in ref. 18, suggesting that reconstructed T(Δ₄₇) of paleosol carbonate in these samples largely represents MAST (see SI Appendix, Supplemental note 1 for detail), rather than summer air temperature as originally described (18). Soil water δ¹⁸O_sw values were calculated from the updated T(Δ₄₇) values and soil carbonate δ¹⁸O_c values (Dataset S1) and shown in SI Appendix, Fig. S5.

Change Point Analysis.

To identify where the mean of the T(Δ₄₇) signal changes most significantly, change points in the mean for the T(Δ₄₇) record between 8 and 5 Ma were calculated using the findchangepts function in MATLAB ver. R2021a (SI Appendix, Fig. S5).

Numerical Climate Modeling.

We used the Community Earth System Model version 1.2 (CESM1.2) to generate the different hydroclimate responses to changing atmospheric pCO₂. In this study, the modules of the atmosphere (CAM4), land (CLM4), river runoff, ocean (POP2), and sea ice are coupled and interact with each other through a coupler. CAM4 operates at a horizontal resolution of f19 (~1.9° × 2.5°) with 26 vertical levels. CLM4 operates at the same horizontal resolution of f19. POP2 employs a gx1v6 grid, which has 384 grid points in the meridional and 320 grid points in the zonal directions, and 60 vertical levels. The sea ice module employs the same horizontal grid as POP2. Three experiments with pCO₂ fixed at 400, 284, and 142 ppm are conducted with other boundary conditions (e.g., land-sea distribution, global topography, vegetation pattern, ice sheets) kept the same as their PI values. Each of these three experiments, termed PI_400, PI_284, and PI_142, is integrated for 800 model years restarted from an existing 1,000-y control run published by Yang et al. (65). The relatively small imbalance in the top of the atmosphere net radiation balance (0.11 W m⁻² for PI_400, −0.02 W m⁻² for PI_284, and −0.27 W m⁻² for PI_142) for the last 100 y and the surface air temperatures trend suggest the surface climate state is approaching equilibrium (SI Appendix, Fig. S12). The computed climatological means from the last 50 model years are analyzed.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data and Materials Availability

All study data are included in the article, SI Appendix and Datasets S1–S3.