Warming-induced northwestward migration of the East Asian monsoon rain belt from the Last Glacial Maximum to the mid-Holocene

Significance

The southward displacement of the East Asian monsoon rain belt heightens concerns over the warming-induced drying of northern China. Paleovegetation change on the Chinese Loess Plateau provides insights into future climate changes. We find that the spatial distribution of C₄ plant biomass is a robust analog for the monsoon rain belt, which migrated at least 300 km to the northwest from the cold Last Glacial Maximum (∼19 ka) to the warm Holocene (∼4 ka). These results strongly support the idea that the Earth's thermal equator will move northward in a warmer world, and that the observed southward migration of the monsoon rain belt over the last few decades is transient and northern China will eventually become wet as global warming advances.

Abstract

Glacial–interglacial changes in the distribution of C₃/C₄ vegetation on the Chinese Loess Plateau have been related to East Asian summer monsoon intensity and position, and could provide insights into future changes caused by global warming. Here, we present δ¹³C records of bulk organic matter since the Last Glacial Maximum (LGM) from 21 loess sections across the Loess Plateau. The δ¹³C values (range: –25‰ to –16‰) increased gradually both from the LGM to the mid-Holocene in each section and from northwest to southeast in each time interval. During the LGM, C₄ biomass increased from <5% in the northwest to 10–20% in the southeast, while during the mid-Holocene C₄ vegetation increased throughout the Plateau, with estimated biomass increasing from 10% to 20% in the northwest to >40% in the southeast. The spatial pattern of C₄ biomass in both the LGM and the mid-Holocene closely resembles that of modern warm-season precipitation, and thus can serve as a robust analog for the contemporary East Asian summer monsoon rain belt. Using the 10–20% isolines for C₄ biomass in the cold LGM as a reference, we derived a minimum 300-km northwestward migration of the monsoon rain belt for the warm Holocene. Our results strongly support the prediction that Earth's thermal equator will move northward in a warmer world. The southward displacement of the monsoon rain belt and the drying trend observed during the last few decades in northern China will soon reverse as global warming continues.

The East Asian summer monsoon plays a crucial role in interhemispheric heat and moisture transport and serves as the main moisture supply for East Asia (1). In recent years, the impact of global warming on the East Asian monsoon has been the subject of intense investigation (2), because even a minor change in monsoon intensity can have a profound effect on the lives of hundreds of millions of people. Several authors (3) argue that meridional asymmetries caused by prominent warming between 45° N and 60° N, compared with the tropics, induces a weakened meridional thermal contrast over eastern Asia and may explain the southward shift in the monsoon rainfall belt (4–7), with more droughts in northern China countered by more floods in southern China, as observed since the 1970s. This is consistent with the prediction of Held and Soden (8) that Earth's dry regions will become drier, and its wet regions wetter, with global warming. Broecker and Putnam (9), however, argue that an increased interhemispheric temperature contrast would tend to shift the thermal equator northward in a warmer world, instead leading to a northward shift in the East Asian summer monsoon rainfall belt and increased precipitation in northern China. It remains unclear whether the recent drying is transient, and linked to El Niño Southern Oscillation (ENSO) or other large-scale variability, or if it signals the start of a long-term drying trend induced by global warming.

The Chinese Loess Plateau is located in the marginal zone of the summer monsoon and is characterized by a steep climatic gradient, where the spatiotemporal change of C₃/C₄ vegetation is closely related to the summer monsoon intensity (10–17). The warming interval from the Last Glacial Maximum (LGM) to the Holocene offers a useful test for future hydroclimatic changes (18). Consequently, we examined spatial changes in C₄ plant biomass for the LGM and Holocene based on δ¹³C records of bulk organic matter across the Loess Plateau, with the aim of estimating the shift of the monsoon rain belt associated with past global warming as well as predicting future hydroclimatic trends.

Spatial Changes in C₄ Plant Abundance During the LGM and Holocene

The modern climate of East Asia is characterized by seasonal alternations of a wet, warm summer monsoon and a dry, cold winter monsoon (Fig. 1). In the Chinese Loess Plateau, mean annual temperature increases from ∼7 to ∼13 °C, and mean annual precipitation from ∼250 to ∼600 mm, from northwest to southeast (Fig. 1), with ∼60–80% of the precipitation concentrated in the summer season. Across these large gradients on the Loess Plateau, the alternation of loess (L) and soils (S) documents large-scale oscillations between glacial and interglacial conditions (19–21). We logged and sampled 21 sections (Fig. 1) to characterize late Quaternary changes in C₃/C₄ vegetation as a proxy for the intensity and position of the East Asian Monsoon rain belt across the Loess Plateau.

Fig. 1.

Chinese Loess Plateau, prevailing monsoon circulation, and modern climatic gradients. Solid circles show locations of 21 sites investigated in this study, and solid triangles denote locations of 4 additional sites previously studied by Liu et al. (14). Arrows in Inset map indicate the direction of the winter and summer monsoonal winds. The isohyets (in millimeters; gray lines) and isotherms (in degrees Celsius; dashed lines) are values averaged over 51 y (1951–2001).

All sections consist of soil unit S₀ and the upper part of loess unit L₁ (Fig. 2 and Fig. S1). The Holocene soil (S₀), overlain by modern topsoil, is dark in color because of its relatively high organic matter content. Loess unit L₁, yellowish in color and massive in structure, was deposited during the last glacial period. L₁ can generally be divided into five subunits termed L_1-1, L_1-2, L_1-3, L_1-4, and L_1-5 (22, 24). L_1-2 and L_1-4 are weakly developed soils, and the other subunits are typical loess horizons. Previous studies (19, 20, 22, 25) demonstrate that (i) L_1-1 was deposited in marine isotope stage (MIS) 2 (∼27–11 ka), which includes the LGM (∼26.5–19 ka) (26); (ii) S₀ was deposited in the early-mid–Holocene (∼11–3 ka), which includes the Holocene Thermal Maximum (HTM) (∼11–5 ka) (27); and (iii) L_1-2 was deposited in late MIS 3 (∼38–27 ka). To ensure that we used a complete cold–warm cycle for C₃/C₄ vegetation reconstruction, almost all of the sections were sampled down to loess unit L_1-2.

Fig. 2.

Stratigraphic column, median grain size (Md), magnetic susceptibility (SUS), and δ¹³C of bulk organic matter for the 21 loess sections (Fig. 1) across the Chinese Loess Plateau, and correlation with a stacked Chinese loess grain-size record [Chinese Loess Millennial-Scale Oscillation Stack (CHILOMOS) (22)] and a stacked benthic δ¹⁸O record (23). The stratigraphic positions of the selected samples used in Figs. 4 and 5 are indicated by solid triangles in each section. Median calibrated radiocarbon ages of the bulk organic fraction for 10 of those selected samples are labeled beside the triangles. Marine isotope stages (MIS) 1–3 are labeled. The shaded zones indicate the Holocene soil (S₀). The gray lines indicate the top of MIS 3 (L_1-2).

Fig. S1.

Correlation of lithostratigraphy, median grain size (Md), magnetic susceptibility (SUS), and δ¹³C of bulk organic matter between the Baoji, Lantian, and Xingyang sections. Note (i) the stratigraphic contrast between soil unit S₀ (Holocene) and loess unit L_1-1 [Last Glacial Maximum (LGM)] in each section, and (ii) the similarity in the structure of Md, SUS, and δ¹³C curves.

In general, soil units S₀ (Holocene) and L_1-2 (late MIS 3) are characterized by higher magnetic susceptibility values and finer grain sizes compared with the LGM loess unit L_1-1, and there is a strong similarity in the structure of the grain size and magnetic susceptibility curves between the different loess sections (Fig. 2). The correlation of lithostratigraphy, grain size, and magnetic susceptibility (Fig. 2 and Fig. S1) between sections demonstrates the remarkable stratigraphic and spatial continuity of the loess deposits at orbital or even millennial timescales.

δ¹³C Contour Maps.

Plants use two principal biosynthetic pathways to fix carbon, C₃ and C₄, which have distinct carbon isotope signatures (28, 29). Modern surveys (30) demonstrate that, in the loess region of northern China, C₃ plants have δ¹³C values ranging from –21.7‰ to –30‰ with a mean of –26.7‰; and that C₄ plants have δ¹³C values ranging from –10‰ to –15.8‰ with a mean of –12.8‰. In all of the sections, the δ¹³C values of soil organic matter fall within the range of –25‰ to –21‰ for the LGM loess unit and within the range of –23‰ to –16‰ for the Holocene soil. From the LGM to the Holocene, the δ¹³C values increased consistently at each site (Fig. 2) and they exhibit a good correlation with the grain size and magnetic susceptibility records. Given that the L_1-1 loess unit has a low organic matter content (<0.5%) (31), the effects of percolation of soluble organic substances (32) and soil texture (33, 34) on the observed glacial–interglacial δ¹³C pattern (Fig. 2) need to be carefully evaluated before definitively linking the δ¹³C records to C₃/C₄ vegetation change.

Radiocarbon (¹⁴C) measurements of bulk organic matter and humin fractions from five sections yielded ages of 4595−3395 cal y B.P. and 4730−3860 cal y B.P., respectively, for the finest-grained part of S₀, and 23,320−16,135 cal y B.P. and 23,930−16,370 cal y B.P., respectively, for the coarsest part of L_1-1 (Fig. 2 and Table 1). The radiocarbon ages of S₀ and L_1-1 overlap with or are slightly younger than the dates of HTM (∼11–5 ka) and LGM (∼26.5–19 ka), respectively. Because measured ¹⁴C ages of soil organic matter are always younger than the true age of deposition, due to addition of younger organic matter through rootlet penetration, bioturbation, and percolation of soluble organic substances (32, 35), our radiocarbon dates are compatible with the widely accepted assumption that the coarsest interval of L_1-1 represents the coldest interval of the LGM, whereas the finest interval of S₀ represents the warmest interval in the Holocene (22, 24).

Table 1.

Accelerator mass spectrometry (AMS) radiocarbon ages for bulk organic and humin fractions from five sections across the Chinese Loess Plateau

Section	Unit	Depth, m	Soil fraction	14C age, y B.P.	2σ cal age, y B.P.	Median cal age, y B.P.	Beta no.
Jingtai	S0	0.55	Bulk sediment	3360 ± 30	3650‒3555	3605	417149
			Humin	3570 ± 30	3935‒3825	3870	414142
	L1-1	4.30	Bulk sediment	13,410 ± 40	16,295‒15,960	16,135	414501
			Humin	19,880 ± 60	24,140‒23,700	23,930	414502
Weiyuan	S0	0.95	Bulk sediment	3380 ± 30	3695‒3565	3625	414153
			Humin	4190 ± 30	4765‒4620	4730	414154
	L1-1	4.05	Bulk sediment	19,370 ± 60	23,570‒23,060	23,320	414155
			Humin	16,990 ± 60	20,680‒20,280	20,490	414156
Qingyang	S0	1.25	Bulk sediment	3170 ± 30	3450‒3345	3395	417151
			Humin	3690 ± 30	4095‒3960	4035	414146
	L1-1	4.00	Bulk sediment	15,020 ± 40	18,405‒18,070	18,255	414503
			Humin	17,470 ± 50	21,315‒20,880	21,095	414504
Lantian	S0	1.15	Bulk sediment	4090 ± 30	4650‒4515	4595	414157
			Humin	3750 ± 30	4160‒4065	4115	415474
	L1-1	2.65	Bulk sediment	15,420 ± 50	18,805‒18,565	18,690	414158
			Humin	13,590 ± 40	16,580‒16,195	16,370	415475
Xingyang	S0	1.25	Bulk sediment	3410 ± 30	3720‒3575	3660	417150
			Humin	3560 ± 30	3930‒3820	3860	415476
	L1-1	4.65	Bulk sediment	13,740 ± 40	16,835‒16,365	16,595	414160
			Humin	13,890 ± 50	17,045‒16,580	16,830	415477

At each site, the samples dated were selected from the finest-grained part of the Holocene soil (S₀) and the coarsest part of the LGM loess unit (L_1-1). The stratigraphic positions of the samples are indicated in Fig. 2. The radiocarbon dates yield an average of ∼3775 cal y B.P. (bulk sediment) and ∼4120 cal y B.P. (humin) for S₀, and ∼18,600 cal y B.P. (bulk sediment) and ∼19,740 cal y B.P. (humin) for L_1-1.

It should be noted that the ¹⁴C ages of bulk organic matter for most samples, especially those from S₀, are ∼15% younger on average compared with the ages obtained from the humin fraction (the stable and insoluble fraction of soil humic substances), indicating a moderate contamination by younger soluble organic substances. However, an average age difference of ∼15 ka between S₀ (average, ∼4 ka) and L_1-1 (average, ∼19 ka), as indicated by the dates for both bulk organic carbon and more stable humin fractions (Table 1), demonstrates a negligible effect of translocation of organic matter from S₀ into the underlying L_1-1. This may result from the fact that Chinese loess has a strong adsorption capacity due to the large clay (<2 μm)−silt (2−50 μm) fraction (>50%; Fig. 3C), which minimizes the potential for contamination by percolation of soluble organic substances (36, 37).

Fig. 3.

Total organic carbon (TOC) content (A), δ¹³C of soil organic matter (B), and weight percentage data (C) for different size fractions in loess (L_1-1) and soil (S₀) samples. TOC and weight percentage were measured on a carbonate-free basis.

Size fractionation results show that (i) soil organic matter is mainly concentrated in clay (<2 μm) and fine silt (2−20 μm) fractions (Fig. 3A), and (ii) δ¹³C values fluctuate without significant size dependence across the observed size range (Fig. 3B), except for the relatively low δ¹³C values for the sand fraction of the Holocene soil samples. However, the effect of this fraction on the δ¹³C of bulk samples is very limited due to the low content of both the sand fraction (Fig. 3C) and its associated organic matter (Fig. 3A). In addition, the δ¹³C value of a specific size fraction in L_1-1 is generally lower than its counterpart in S₀ (Fig. 3B), exhibiting the same pattern as observed for the bulk δ¹³C record (Fig. 2). In this context, the glacial–interglacial δ¹³C pattern (Fig. 2) faithfully mirrors C₃/C₄ vegetation change.

To plot the δ¹³C contour maps, we selected the δ¹³C value of a sample from within the interval of coarsest grain size of L_1-1 for the LGM, and that of a sample from within the interval of finest grain size of S₀ for the mid-Holocene. The δ¹³C contour maps were constructed using the kriging algorithm in the Surfer software package. The δ¹³C isolines exhibit a northeast–southwest trend in both the LGM and the mid-Holocene (Fig. 4), which is generally consistent with the present-day climatic pattern, i.e., a northeast–southwest trend for both the modern annual isohyets (Fig. 4 A and C) and isotherms (Fig. 4 B and D). From northwest to southeast, the δ¹³C values increase from –24‰ to –22‰ during the LGM and from –22.5‰ to –17.5‰ during the mid-Holocene.

Fig. 4.

δ¹³C (‰) contour maps for the Last Glacial Maximum (LGM) (A and B) and the mid-Holocene (C and D), together with modern mean annual precipitation (A and C) and temperature (B and D) over 51 y (1951–2001). The contour maps were constructed using δ¹³C data from the 21 sites in this study (solid circles; Figs. 1 and 2) and four additional sites in Liu et al. (14) (solid triangles; Fig. 1). All of the δ¹³C data used in this paper were obtained from the same laboratory by identical methods, thereby minimizing interlaboratory variability.

C₄ Abundance Contour Maps.

To estimate the C₃/C₄ biomass, it is crucial to determine end-member δ¹³C values for C₃ and C₄ plants in the LGM and mid-Holocene. In so doing, we considered the effects of the δ¹³C of atmospheric CO₂ (δ¹³C_atm), precipitation, and temperature change on the δ¹³C values of C₃ and C₄ plants (Table S1). However, we ignored the effect of changes in atmospheric CO₂ concentration because the relationship between CO₂ concentration and δ¹³C of C₃ plants on the Loess Plateau is unknown. After correcting for these factors, and for organic matter degradation, the end-member δ¹³C values of soil organic matter for C₃ (δ¹³C_C3) and C₄ plants (δ¹³C_C4) were obtained for the LGM and mid-Holocene (Table S1). C₄ plant abundance was estimated by applying the measured δ¹³C values to an isotope mass-balance equation: C₄(%) = [(δ¹³C – δ¹³C_C3)/(δ¹³C_C4 – δ¹³C_C3)] × 100.

Table S1.

Calculation of end-member δ¹³C values of soil organic matter for C₃ (δ¹³C_C3) and C₄ vegetation (δ¹³C_C4) for the mid-Holocene and Last Glacial Maximum (LGM)

Factor	δ13CC3, ‰	δ13CC4, ‰	Calculation basis and data sources
Mid-Holocene*
Today's vegetation	−26.7	−12.8	Ref. 30
δ13Catm correction (+1.7‰)	−25.0	−11.1	δ13C of atmospheric CO2: −8.06‰ for the present (Mauna Loa Observatory, 2000; ftp://aftp.cmdl.noaa.gov/data/trace_gases/co2c13/flask/surface/co2c13_mlo_surface-flask_1_sil_month.txt); −6.33‰ for the mid-Holocene (53)
Degradation correction (+1.0‰)	−24.0	−10.1	Refs. 17 and 54
LGM
Today's vegetation	−26.7	−12.8	Ref. 30
δ13Catm correction (+1.6‰)	−25.1	−11.2	δ13C of atmospheric CO2: −8.06‰ for the present (Mauna Loa Observatory, 2000; ftp://aftp.cmdl.noaa.gov/data/trace_gases/co2c13/flask/surface/co2c13_mlo_surface-flask_1_sil_month.txt); −6.46‰ for the LGM (53)
Precipitation correction (C3: +0.6‰)	−24.5	−11.2	∼150 mm lower in the LGM than at the present (55, 56); −0.40‰/100 mm for C3 plants and a negligible coefficient for C4 plants (30)
Temperature correction (C3: −0.6‰)	−25.1	−11.2	6 °C lower in the LGM than at the present (56); 0.104‰/°C for C3 plants and no significant correlation for C4 plants (57)
Degradation effect (+1.0‰)	−24.1	−10.2	Refs. 17 and 54

^*The climatic conditions in the mid-Holocene are assumed to be similar to the present, and thus the precipitation and temperature corrections for the mid-Holocene are ignored.

The LGM and mid-Holocene share two common features in the C₄ biomass contour maps (Fig. 5). First, the isolines of C₄ biomass exhibit a northeast–southwest zonal distribution pattern. Second, C₄ biomass increases consistently from northwest to southeast. These similarities demonstrate that the C₄ abundance in both the LGM and mid-Holocene was controlled by the same environmental factor(s). However, two differences between the two time intervals are noteworthy. First, C₄ vegetation increased considerably throughout the Loess Plateau in the mid-Holocene compared with the LGM, with an increase of ∼15% in the northwest and ∼25% in the southeast. Second, a greater spatial change in C₄ biomass is evident in the mid-Holocene (from 10% to 20% in the northwest to >40% in the southeast) than in the LGM (from <5% in the northwest to 10–20% in the southeast). These differences indicate that the environmental factor(s) controlling C₄ abundance varied considerably between the LGM and Holocene.

Fig. 5.

Contour maps of estimated C₄ biomass (%) for the Last Glacial Maximum (LGM) (A and B) and the mid-Holocene (C and D), together with modern mean warm-season precipitation (A and C) and temperature (B and D) (May to September) over 51 y (1951–2001). The C₄ biomass was calculated using δ¹³C data from the 21 sites in this study (solid circles; Figs. 1 and 2) and four additional sites in Liu et al. (14) (solid triangles; Fig. 1). Note the migration of the 10–20% isolines for C₄ biomass (pink lines) between the LGM and the mid-Holocene.

Factors Controlling C₄ Abundance on the Loess Plateau.

Plant biogeographical studies have shown that lower atmospheric pCO₂, higher growing season temperature, and enhanced summer precipitation favor C₄ over C₃ plants (38). Given that atmospheric CO₂ concentrations rose from ∼180 ppmv in the LGM to ∼260 ppmv in the mid-Holocene (39), the significant increase in C₄ abundance during this interval (Figs. 2 and 5) cannot be explained by pCO₂ change. On the Chinese Loess Plateau, both growing season temperature and precipitation are higher in the southeast than in the northwest (Fig. 5), all favoring increased C₄ vegetation in the southeast. However, the relative importance of temperature and precipitation for C₄ vegetation cannot be disentangled due to the seasonal synchrony of rainfall and temperature in the East Asian monsoon region (Fig. 5). Because warm-season precipitation is a defining feature of the summer monsoon circulation (16), it is clear that the northeast–southwest zonal distribution pattern of C₄ biomass on the Loess Plateau can serve as a robust analog for the East Asian summer monsoon rain belt.

Migration of the East Asian Summer Monsoon Rain Belt from the LGM to the Mid-Holocene

The similarities in the spatial pattern of C₄ abundance between the LGM and mid-Holocene (Fig. 5) demonstrate a similar pattern of the summer monsoon rain belt during the two time intervals, i.e., a northeast–southwest trend for the monsoon isohyets through the recent cold–warm cycle. It is especially striking that the C₄ percentage isolines of a specific value exhibit a significant northwest–southeast migration between the LGM and mid-Holocene (Fig. 5). For example, the 10–20% isolines for C₄ biomass in the southeastern part of the Plateau during the LGM moved to the northwestern part during the mid-Holocene, indicating a major northwestward advance of the summer monsoon rain belt from the LGM to the mid-Holocene. Using the plot of 10–20% C₄ biomass as a reference (Fig. 6), we estimate a northwesterly monsoon rain belt advance of ∼300 km for the warm Holocene compared with the cold LGM.

Fig. 6.

Migration of the East Asian summer monsoon rain belt (purple dashed lines) between the Last Glacial Maximum (LGM) and the mid-Holocene as reconstructed using the 10–20% isolines for C₄ biomass shown in Fig. 5.

It should be noted that the decline of atmospheric CO₂ level in the LGM would have caused increased δ¹³C values of C₃ plants, because modern observations (40, 41) reveal a wide range of coefficients, from −0.019‰/100 ppmv to −2.7‰/100 ppmv, depending on targeted species and CO₂ concentrations. In view of an ∼80-ppmv increase in atmospheric CO₂ levels in the mid-Holocene relative to the LGM (39), the end-member δ¹³C of C₃ plants for the LGM should be higher than that used for the calculation of C₄ biomass (Table S1). Thus, the C₄ percentages estimated for the LGM (Fig. 5 A and B) should be regarded as upper limits. With regard to the mid-Holocene (S₀), the C₄ biomass may be overestimated because the microbial effect on ¹³C enrichment appears to be more pronounced for C₃- than for C₄-derived organic matter (42, 43). However, this effect may be counterbalanced by the addition of modern organic matter, characterized by relatively low δ¹³C values, to S₀ (Fig. 2), which is indicated by the young ¹⁴C ages obtained from both the bulk organic and humin fractions (Table 1). In this context, the estimated ∼300-km rain belt advance can be regarded as a minimum for the mid-Holocene (Fig. 6).

Paleotemperature reconstructions demonstrate that from the LGM to the Holocene (preindustrial), global mean sea surface temperature increased by 0.7–2.7 °C (44, 45), whereas global mean surface air temperature increased by as much as 3–8 °C (45), indicating a greater warming over land areas than over oceans. Our results demonstrate a northwestward advance of the East Asian summer monsoon rain belt from the cold LGM to the warm Holocene. This may be explained by the following physical process: the decay of continental ice in the Northern Hemisphere by the HTM, as well as the vast continent itself (warming more rapidly than the oceans), caused a northward shift of Earth's thermal equator, thus driving the northward migration of the Asian monsoon rain belt (9). An additional process could be the rise of global sea level in the Holocene, which led to a northwesterly transgression of the Western Pacific marginal seas (46), thereby facilitating the northwestward penetration of the monsoon rain belt.

Our study strongly supports the prediction of Broecker and Putnam (9) that monsoonal Asia will become wetter in a warmer world. The observed drying trend in northern China since the 1970s is likely to be caused by increased winter and spring snow ice cover in the Tibetan Plateau (47) and by an ENSO-like mode of sea surface temperature (48). In this context, northern China is expected eventually to become wet as global warming continues.

Materials and Methods

A total of 2,383 samples were collected at a 5-cm interval. For stratigraphic correlation, grain size and bulk magnetic susceptibility were measured on all samples using a SALD-3001 laser diffraction particle analyzer and a Bartington Instruments MS2 magnetic susceptibility meter. The grain-size analytical procedures were as detailed by Ding et al. (49).

A total of 590 samples were selected for δ¹³C analysis of bulk organic matter. Sample splits (∼2 g) were first screened for modern rootlets and then digested for 24 h in 1 M HCl at room temperature to remove inorganic carbonate. The samples were then washed to pH >4 with distilled water and dried cryogenically at −80 °C. The dried samples (∼500 mg) were combusted for over 4 h at 850 °C in evacuated sealed quartz tubes in the presence of 1 g of Cu, 1 g of CuO, and a Pt wire. The CO₂ was purified and isolated by cryogenic distillation for isotopic analysis. Carbon isotopic composition of CO₂ was then determined using a MAT-251 gas mass spectrometer at the Institute of Earth Environment, Chinese Academy of Sciences (CAS). The δ¹³C results are reported in per mil units (‰) relative to Vienna Peedee belemnite (VPDB) standard with an error of less than 0.2‰.

Ten samples were selected from S₀ and L_1-1 at five widely separated sites for the measurement of accelerator mass spectrometry (AMS) ¹⁴C ages for both bulk organic carbon (acid wash treatment) and humin (acid–alkali–acid treatment) fractions. All samples were pretreated and analyzed by Beta Analytic. Calibration of ¹⁴C dates was done in Calib Rev 7.0.4 (calib.qub.ac.uk/calib/calib.html) (50) using the IntCal13 curve (51).

Eight samples from the L_1-1−S₀ couplet at four sites were separated into sand (>50 μm), coarse silt (20−50 μm), fine silt (2−20 μm), and clay (<2 μm) fractions according to standard sieving and pipette methods (52). Bulk samples (20−30 g) were first screened for modern rootlets and decarbonated as described above, and then dispersed in 300 mL of distilled water with 10 mL of 0.05 M (NaPO₃)₆ and treated ultrasonically. The sand (>50 μm) fraction was separated by wet sieving, and the resulting <50-μm suspension was separated into 20- to 50-μm, 2- to 20-μm, and <2-μm fractions by gravity sedimentation (52). Particle size separates were then freeze-dried, weighed to obtain a mass for each fraction, ground, and analyzed to determine δ¹³C and total organic carbon (TOC) content. TOC content of the samples (200 mg for each) were measured by high-temperature combustion (950 °C) using an Elementar rapid CS CUBE.