Introduction

Peatlands cover only 3% of the global land but contain 500–700 Gt of carbon, accounting for around 30% of global soil carbon stocks1,2. They act as key carbon reservoirs that have been sequestering atmospheric carbon in peat deposits during the past millennia3,4,5. With intensified climate change and anthropogenic activity, peatlands are experiencing substantial ecological shifts, with widespread encroachment of woody plants such as shrubs and trees observed in many peatlands6,7,8,9,10, driven by warming and lower water tables that favor shrub and tree root growth11. Woody plant expansion not only reduces Sphagnum mosses and herbaceous plants but also influences peat microbial metabolism6,12, affecting carbon accumulation dynamics and the long-term balance between carbon storage and release. Yet, the long-term impacts of woody plant encroachment on carbon storage remain uncertain - a critical gap, as projections suggest intensified woody expansion in the coming decades9,11. Understanding how woody plant expansion shapes microbial communities, metabolic activity, and carbon dynamics in peatlands is therefore essential.

Woody plant expansion influences peatland microbes by altering root exudates and litter chemistry6,12. As shown in field manipulation experiments, woody plant root exudates can stimulate peat microbial growth and production of carbon-degrading enzymes6, causing ancient peat decomposition and increasing CO2 emissions from peatlands13. A shift from Sphagnum to shrub has been shown to reduce bacterial diversity and increase dominance of a few keystone taxa, especially Acidobacteriota, suggesting an unstable belowground microbial community and potential degradation of peatlands14. In contrast, polyphenolic-rich woody litter suppresses microbial metabolism, slowing microbial decomposition and potentially enhancing carbon retention12,15. These contrasting effects suggest woody plant expansion has a dual and competing impact on microbial processes, necessitating further research to clarify its effect on peatland carbon stocks. However, studies of woody plant expansion typically focus on short-term impacts rather than the centennial-to-millennial scales essential for understanding long-term peatland carbon dynamics.

Peatlands have initiated and stored huge amounts of carbon since the Late Pleistocene, with peat deposits providing valuable records of long-term carbon accumulation as a balance of plant production and microbial decomposition4,5,16. Understanding these long-term carbon dynamics is crucial for predicting the future stability of these significant carbon stocks16,17,18,19. In peat deposits, plant macrofossils have been widely used to reconstruct historical vegetation changes20 and reveal past shifts to woody plants that occurred in response to warm and dry periods8,21,22. Recent research documents ongoing woody expansion in circum-Arctic peatlands over the last 8000 years, often as a consequence of drier conditions23. However, linking these vegetation shifts with microbial activity and carbon accumulation over millennia remains challenging, as few records capture the integrated dynamics of plant community, microbial activity, and carbon storage across such long timescales.

In this study, we employ an integrated approach using multiple proxies from a southern Chinese peatland—Zhaogongting (ZGT) (Fig. 1) to explore how woody plant expansion and microbial interactions regulated carbon accumulation from the Late Pleistocene to the Holocene. Using plant macrofossils, we reconstructed plant community changes and identified significant mid-Holocene woody expansion events under drying conditions. Additionally, we analyzed bacterial lipid biomarkers, hopanoids, and their compound-specific carbon (δ13C) and hydrogen (δ2H) isotope compositions, as well as fungal-derived fatty acid biomarkers. The δ13C and δ2H isotopic profiles of these microbial lipids allow us to track shifts in microbial carbon cycling and metabolism24,25,26,27. Additionally, we determined the peat’s carbon accumulation history and used Fourier Transform Infrared Spectroscopy (FTIR) and bulk geochemistry to analyze its organic matter composition, representing both its source and degradation, creating a framework that directly links vegetation, microbial activity, peat organic matter composition, and carbon accumulation dynamics. Our results demonstrate that woody plant expansion during the Holocene triggered significant fluctuations in bacterial and fungal biomass, carbon metabolism, and peat organic composition, with variations in heterotrophic activity closely tied to woody-driven changes in peat organic matter composition. The combination of transitions to recalcitrant organic matter pools and suppressed heterotrophic metabolism, resulted in the highest carbon accumulation rates observed from the Late Pleistocene to the Holocene, highlighting the critical role of microbial responses to changing plant community in enhancing peatland resilience and protecting long-term carbon storage under climate stress.

Fig. 1: Site location and field photograph of ZGT peatland.
figure 1

The public elevation data are obtained from the USGS EROS Archive - Shuttle Radar Topography Mission (SRTM), provided by USGS/NASA (https://earthexplorer.usgs.gov). Field photograph taken in September 2016.

Full size image

Results and discussion

Woody plant expansion during the Holocene

Plant macrofossils from the ZGT peat core reveal significant shifts in vegetation since the Late Pleistocene. Herbaceous plants and Sphagnum mosses dominated from the Late Pleistocene to the Early Holocene, accounting for roughly 80% of the total community, while woody plants comprised only about 20% (Fig. 2a). However, during the early mid-Holocene (8–6 ka), woody plant proportions increased dramatically in three phases, reaching nearly 90% (Fig. 2a). This interval also contains abundant Sphagnum (Fig. 2a), including two species, S. palustre and S. subsecundum, indicating that this had not become a forested swamp but a Sphagnum bog with a significant component of wood-forming shrubs and/or trees. After 6 ka, the proportion of woody plants stabilized back to around 20%, with herbaceous plants once again becoming dominant (Fig. 2a). Additional paleoecological studies from other peatlands in central-southern China, including Shuizhuyang, Daping, Shiwangutian, Dajiuhu, and Hongyuan, report increases in tree pollen percentages around 8 ka (Supplementary Fig. S1). These pollen records most likely reflect regional upland vegetation dynamics rather than local peat-forming communities within the peatlands. Therefore, while informative as regional background, we do not interpret them as evidence for a regionally synchronous expansion of woody plants across central-southern Chinese peatlands.

Fig. 2: Holocene shifts in vegetation, microorganisms, peat organic composition, and carbon accumulation in ZGT peatland.
figure 2

a Proportion of plant macrofossils. b δ13C values of C29 and C31 n-alkanes. c Hopanoid flux35. d Total concentration of bacterial C27β, C29βα, C29ββ, C30ββ, C31αβ, and C31ββ hopanes. e Concentration of fungal C18:2ω6,9 fatty acid. f δ13C of C31αβ hopane. g The carbon isotope offset (εhopane/bulk) between δ13C of C31αβ hopane and bulk organic carbon. h δ2H of C31αβ hopane. i The hydrogen isotope fractionation (εhopane/water) between C31αβ hopane and source precipitation water. j Proportion of carbohydrates. k Proportion of aromatics. l Total organic carbon content35. m Apparent carbon accumulation rate (aCAR)35. n Bulk density35. The gray shading marks the 8–6 ka period of woody plant expansion.

Full size image

Our interpretation of woody plant expansion at the ZGT peatland is based on plant macrofossil evidence. However, our plant macrofossil reconstruction does not extend to the present due to data limitations. The uppermost peat layers are highly decomposed and compacted, making it difficult to recover well-preserved plant remains for reliable identification. Additionally, potential disturbances from modern land use may have affected the integrity of the surface peat record. Despite this limitation, our field surveys indicate that the modern ZGT peatland features a mix of shrubs (Camellia caudata, Rhododendron spp.), sedges (Carex spp.), and mosses (Sphagnum palustre, Polytrichum commune) (Fig. 1), suggesting that woody vegetation remains a component of the ecosystem but has not dominated as extensively as during the mid-Holocene expansion.

The mid-Holocene expansion of woody vegetation in the ZGT peatland appears to be tied to both warming and drying trends. Temperature reconstructions from the nearby Shuizhuyang peatland and Huguangyan Maar Lake, derived from branched glycerol dialkyl glycerol tetraethers, indicate a warming around 8 ka corresponding with the Holocene Thermal Maximum28,29. This warming trend, consistent with global Holocene patterns30, likely supported woody plant expansion by promoting growth in warmer conditions, similar to the processes observed in circum-Arctic peatlands starting around 8 ka23. Modern studies also demonstrate that warming accelerates shrub and tree root growth, enabling woody plants to outcompete herbaceous vegetation in peatlands; for example, every degree increase in peat soil temperature has been shown to increase shrub fine root growth by 130%, providing shrubs with greater access to nutrients and water11.

Local hydroclimatic changes also played a key role in driving woody plant expansion. In the ZGT peat core, δ13C values of C29 and C31 n-alkanes show variability during the mid-Holocene, increasing by over 2‰ around 7–6 ka (Fig. 2b). However, this δ13C change may result from both local hydrology and vegetation changes. To investigate, we analyzed δ13C values of C29 and C31 n-alkanes from modern woody plants, herbaceous plants, and Sphagnum mosses in ZGT peatland, revealing significant differences among the three plant types, with woody plants having ~2‰ lower δ13C values (Supplementary Fig. S2; Table S1). This suggests that woody expansion cannot account for the elevated δ13C observed in the ZGT peat core. Instead, the elevated n-alkane δ13C values may reflect reduced methanotrophy and/or enhanced water-use efficiency under drier conditions26,31,32,33,34. Our earlier analyses of δ2H values of n-alkanes in the ZGT core show an overall increase from the early to late Holocene, indicating reduced moisture or drier conditions, albeit with particularly profound variability from 8 to 6 ka35. During the same interval, hopanoid flux also increased markedly (Fig. 2c), reflecting lower water tables and drier conditions in the peatland36. This reduction in moisture availability contributed to drier conditions that facilitated woody plant encroachment. This phenomenon has been observed in peatlands in other areas of the world, where woody expansion was driven by drying and general water table drawdown23,37,38. In turn, the establishment of woody plants may further contribute to local drying through enhanced evapotranspiration and well-developed root systems11,39, reinforcing the feedback between vegetation and peatland hydrology.

Microbial biomass and metabolism shifts during woody plant expansion

Hopanoid biomarkers in the ZGT peat core reflect dynamic bacterial responses to woody plant expansion between 8 and 6 ka (Fig. 2). Total concentrations of C27β, C29βα, C29ββ, C30ββ, C31αβ, and C31ββ hopanes (where subscript refers to carbon number and Greek letters refer to stereochemistry at the C-17 and C-21 positions, respectively) initially increased, with a single-point peak around 7.3 ka (Fig. 2d); however, statistical analysis indicates no significant (p = 0.40, t = 0.86, df = 25, n = 27) difference in hopane concentrations between the 8–6 ka interval and other periods. The dominant C31αβ hopane, along with other hopanes such as C29βα and C30ββ, followed similar trends, although others do not suggest a changing bacterial community throughout the entire record (Supplementary Fig. S3). While the isolated peak at 7.3 ka may reflect a short-term increase in bacterial biomass or hopanoid production under drier conditions36,40,41, localized variability or analytical uncertainty cannot be excluded. Intriguingly, aside from the isolated peak at 7.3 ka, hopane concentrations show an overall decline relative to the early stage of the 8 ka interval, suggesting that putatively aerobic bacterial biomass decreased during the period of woody plant expansion (Fig. 2d and Supplementary Fig. S3), despite evidence for persistently dry conditions. This decline in bacterial biomass likely results from the continuous accumulation of lignin-rich woody residues, which reduce resources for bacterial growth and metabolism12.

Similarly, the fungal biomarker C18:2ω6,9 fatty acid (where subscript 18 refers to carbon number, the colon and the number 2 indicate two double bonds, while ω and the numbers 6,9 indicate the positions of the double bonds) in the ZGT peat core exhibited an initial increase followed by a gradual decline during the woody plant expansion period. The concentration of C18:2ω6,9 fatty acid sharply increased around 8 ka (Fig. 2e), indicating a rapid rise in fungal abundance coinciding with the onset of local drying. This response likely reflects the stimulation of fungal growth under drier conditions, due to enhanced soil aeration and increased fungal metabolism41. Intriguingly, after 8 ka, the concentration of C18:2ω6,9 fatty acid gradually declined (Fig. 2e), indicating a decrease in fungal abundance despite the continued presence of dry conditions. This decline likely reflects a fungal adaptation to the progressive accumulation of phenolic-rich woody residues. For example, in shrub-dominated peatlands, fast-growing fungi are found to be suppressed and replaced by slow-growing taxa adapted to phenolic and recalcitrant substrates15. Therefore, our findings further highlight the inhibitory effect of woody plant expansion on fungal growth in peatlands over millennial timescales during the Holocene.

In addition to biomass, the δ13C and δ2H values of bacterial hopanoids in our ZGT core suggest shifts in bacterial carbon sources and metabolism during woody plant expansion. The δ13C of the dominant C31αβ hopane exhibited considerable variation between −24‰ and −32‰ during the 8–6 ka period (Fig. 2f), with similar fluctuations observed for C27β and C30ββ hopanes (Supplementary Fig. S4). In contrast, the δ13C of bulk peat organic carbon (δ13Cbulk) remained stable (Supplementary Fig. S4). As a result, the offset between the δ13C of C31αβ hopane and bulk organic carbon, εhopane/bulk, varied between −4 and 4‰ (Fig. 2g), suggesting a change in bacterial carbon source and/or metabolism during woody plant expansion. This may reflect a shift between 13C-enriched substrates (e.g., sugars) and 13C-depleted substrates (e.g., respired CO2 or methane), and/or a transition between heterotrophy and autotrophy/methanotrophy24,26,42,43.

The δ13C values of C18 fatty acid also fluctuated markedly during the 8–6 ka interval, ranging from −30.8‰ to −24.7‰ (Supplementary Fig. S4). On average, these values (avg. −26.7‰) were ~1.5‰ more enriched than both bulk organic carbon and bacterial hopanoids (avg. −28.0‰), potentially indicating that microbes producing C18 fatty acids preferentially utilized 13C-enriched substrates (e.g., sugars). Given that saturated C18 fatty acids can originate from fungi, bacteria, and higher plants, the observed δ13C values likely reflect a combination of microbial and plant-derived carbon sources. While source attribution remains uncertain, the enriched values could reflect fungal contributions and support the interpretation of altered microbial metabolism during woody plant expansion under drier conditions.

The δ2H values of the bacterial hopane also reveal dynamic metabolic behavior during woody plant expansion, although with different patterns than those of the δ13C values. δ2H values of the C31αβ hopane—for which we obtained the highest resolution record—generally increase from the early to late Holocene, with the major shift occurring from 8 to 6 ka and paralleling trends in n-alkane δ2H values, albeit with a smaller magnitude. However, this trend is punctuated by a dramatic perturbation in the woody interval, with hopane δ2H values initially increasing from −260‰ to −240‰ from 8 to 7 ka before sharply decreasing to −280‰ around 7.0 ka and then rebounding to −250‰ (Fig. 2h). These shifts could relate to changes in the δ2H of source precipitation water, especially during hydroclimatic transitions. To eliminate source water δ2H variability, we calculated the εhopane/water value, representing hydrogen isotope fractionation between bacterial C31αβ hopane and precipitation water. Even after accounting for source water δ2H changes, εhopane/water values still showed considerable variation, rising from −210‰ to −190‰ at the onset of woody plant expansion and then decreasing to around −230‰ at 7.0 ka (Fig. 2i).

Alternatively, the changes in εhopane/water could reflect changes in microbial metabolism pathways, e.g., with autotrophs generally exhibiting lower εlipid/water values than heterotrophs25,44,45,46,47. Consequently, the observed ~40‰ drop in εhopane/water during woody plant expansion could reveal a transient metabolism shift from heterotrophy to greater autotrophy (Supplementary Fig. S5). Differences in biosynthetic pathways add complexity. Bacteria synthesize hopanoids primarily via the methylerythritol phosphate (MEP) pathway, but some heterotrophic bacteria also use the mevalonic acid (MVA) pathway48. These pathways generally lead to lower εlipid/water values than that of n-alkyl lipids synthesized via the acetogenic pathway47. Consistent with this, our εhopane/water values range from −195‰ to −232‰, with an average of −209‰, which is significantly lower than the εlipid/water values of heterotrophic bacterial fatty acids observed in culture (−150‰~+400‰)25,44. However, εlipid/water values also differ between isoprenoid lipids biosynthesized via the MEP and MVA pathways, with lipids produced through the MEP pathway exhibiting greater fractionation and being more 2H-depleted26,47,49. Therefore, the changes in εlipid/water (Fig. 2i) likely document a combination of metabolic (i.e., heterotrophy vs autotrophy) and biosynthetic (MEP vs MVA) processes, and understanding of both is limited. Nonetheless, the variation in hopane δ2H and εhopane/water values between 8 to 6 ka, combined with the variation in hopane δ13C values and the change in hopane abundances and distributions, indicates a dynamic bacterial response to the ZGT vegetation transition.

Microbial metabolism shifts are linked to peat organic composition changes

FTIR spectra reveal peat organic matter composition changes associated with woody plant expansion. Proportions of carbohydrates (related to cellulose and hemi-cellulose) and aromatics (related to lignin and therefore woody tissues) exhibit pronounced fluctuations between 8–6 ka (Fig. 2j, k), corresponding to changes in the relative proportions of woody plants and Sphagnum mosses (Fig. 2a). Peaks in woody plant proportions are associated with higher proportions of aromatics (r = 0.61, p = 0.0265, n = 13) and lower proportions of carbohydrates (r = −0.55, p = 0.0534, n = 13) (Supplementary Figs. S6 and S7a, S7b). FTIR spectra also show higher peaks for lignin and aromatic compounds during the woody plant expansion period compared to non-expansion periods (Supplementary Fig. S8). These findings suggest that vegetation shifts drive peat organic matter towards a more recalcitrant50, aromatic composition during woody plant expansion. This pattern aligns with modern latitudinal trends in peatlands, where tropical and subtropical peatlands, dominated by woody vegetation, show lower carbohydrates and higher aromatic levels than high-latitude peatlands51. Intriguingly, the shift towards greater peat aromaticity (and carbohydrate depletion) precedes the woody interval by several thousand years (Fig. 2j, k), likely driven not only by the vegetation changes during early Holocene (Fig. 2a) but also by warming and drying-induced decomposition52.

The observed changes in peat organic matter recalcitrance during woody plant expansion played a pivotal role in driving bacterial metabolic shifts. Around 7 ka, bacterial εhopane/water values decreased by ~40‰, coinciding with a sharp drop in carbohydrates from 40% to 25% and a > 3% rise in aromatics (Fig. 2i, j, k). Correlation analysis reveals that εhopane/water values are significantly positively correlated with carbohydrates (r = 0.67, p = 0.0043, n = 16) and significantly negatively correlated with aromatics (r = −0.64, p = 0.008, n = 16) (Supplementary Figs. S6 and S7c, S7d), indicating that higher carbohydrate proportions sustain heterotrophy, while elevated aromatics suppress it. In addition, the δ13C values of saturated C18 fatty acid—likely derived from both fungal and bacterial sources—show similar patterns, with significant positive correlation with carbohydrates (r = 0.48, p = 0.0376, n = 19) and negative correlation with aromatics (r = −0.56, p = 0.0127, n = 19) (Supplementary Figs. S7e, S7f). These relationships suggest that microbes contributing to C18 fatty acid production, including fungi, may have adjusted their carbon metabolism in response to peat organic composition changes.

These findings align with broader peatland studies linking increases in organic matter recalcitrance to reduced microbial decomposition and potentially altered metabolism12,15. Labile carbohydrates provide energy-rich substrates critical for heterotrophic metabolism, but their decline, driven by woody plant expansion, limited substrate availability and prompted bacterial adaptation toward autotrophy or mixotrophy53. Similarly, elevated aromatics, particularly woody litter-derived polyphenols, could have further suppressed heterotrophic metabolism by interfering with microbial enzymatic activity54,55. This also explains why we observe no clear trend in bacterial hopanoid abundance, with aerobic conditions that favor bacterial hopanoid-producers countered by organic matter energy limitation. Collectively, our results highlight how woody plant-driven changes in peat organic matter composition constrained microbial heterotrophic metabolism, underscoring the pivotal role of vegetation shifts in regulating peatland microbial dynamics.

Implications for carbon storage in peatlands

Insights into ancient microbial ecology reveal the underlying mechanisms of carbon sequestration during ecological transitions. Total organic carbon (TOC) and apparent carbon accumulation rates (aCAR) significantly increased during the 8–6 ka period, with TOC rising from 20% to nearly 40% (Fig. 2l), and aCAR increasing from below 40 g C/m2/yr to a peak of 146 g C/m2/yr around 7 ka (Fig. 2m). This remarkable increase occurred despite the concurrent warming and drying conditions (Fig. 2b and c), which has been suggested to decrease peatland carbon accumulation through increased organic matter degradation56. Because aCAR is derived from peat accumulation rate, TOC, and bulk density, we further examined whether changes in bulk density contributed to the observed aCAR increase. During the 8–6 ka interval, bulk density mostly varies between 0.43 and 0.66 g/cm3, with a single-point peak at 0.85 g/cm3 at 7 ka (Fig. 2n), coinciding with the highest aCAR value (Fig. 2m). This alignment suggests that increased compaction may have partially contributed to the aCAR spike. However, bulk density during 8–6 ka is similar to the other periods (Fig. 2n), and even lower than last deglacial values (e.g., 0.8–1.0 g/cm3 between 14 and 12 ka), indicating that the bulk density cannot explain the high aCAR during 8–6 ka.

We propose that woody plant expansion and the resulting changes in peat organic matter composition and microbial responses have protected peatland carbon storage at the ZGT peatland during periods of warming and drying. Correlation analyses further support this hypothesis, revealing significant relationships between aCAR and woody plant macrofossils, aromatics, carbohydrates, and microbial biomarker δ13C and εhopane/water values (Supplementary Figs. S6 and S7). These findings highlight three key factors that may modulate the climatic impact on carbon accumulation: vegetation, peat organic matter composition, and microbial metabolism.

First, our findings emphasize the well-established central role of vegetation in shaping peatland carbon dynamics (Fig. 3). A positive correlation between woody macrofossil abundance and aCAR (r = 0.46, p = 0.0437, n = 20) (Supplementary Figs. S6 and S7g) suggests that woody plant expansion directly stimulated carbon input (Fig. 3, Pathway 1). This can be attributed to woody-driven enhanced primary productivity, as woody plants can contribute to higher primary productivity through fast-growth root systems and a greater capacity to fix carbon compared to herbaceous vegetation11,13. Plant primary productivity has been proposed to play a critical role in determining peatland carbon accumulation over long timescales17,57, with the capacity to offset carbon losses induced by warming and drying up to a certain point16.

Fig. 3: Diagram of climate-driven woody plant expansion in peatlands and its effects on peat organic matter composition, microbial metabolism, and carbon accumulation.
figure 3

Three pathways illustrate how woody plant expansion enhances carbon accumulation in peatlands: (1) increasing primary productivity, (2) altering peat organic matter composition by increasing aromatics and reducing carbohydrates, and (3) inhibiting microbial heterotrophic metabolism. Correlation coefficients (r) are shown, with blue arrows indicating positive effects and red arrows indicating negative effects.

Full size image

Second, peat organic matter composition governs its reactivity and therefore carbon accumulation rates (Fig. 3). aCAR in the ZGT peatland exhibits a significant positive correlation with the proportion of aromatics (r = 0.51, p = 0.001, n = 38) and a negative correlation with the proportion of carbohydrates (r = −0.42, p = 0.0083, n = 38) (Supplementary Figs. S6, S7h, S7i), directly linking peat composition to carbon storage. The reduction in labile carbohydrates and increase in recalcitrant aromatics, largely derived from lignin and phenolic compounds in woody residues, indicate a transition toward recalcitrant organic matter pools (Fig. 3, Pathway 2). These differences, originally set by the dominant peat-forming vegetation, are enhanced by the preferential preservation of the more recalcitrant components; in the ZGT peatland, this affects both the woody interval and the underlying horizons, which could have undergone secondary decomposition during drier periods. Such changes create a chemically stable peat matrix dominated by recalcitrant components58,59, offering enhanced resistance to degradation and enabling sustained carbon accumulation during mid-Holocene warm and dry conditions.

Finally, the interplay between peat organic matter composition and microbial metabolism further contributes to the resilience of peatland carbon storage (Fig. 3). The significant negative correlation between aCAR and bacterial metabolic εhopane/water values (r = −0.68, p = 0.0151, n = 12) highlights the regulatory role of microbial communities on carbon accumulation (Supplementary Figs. S6 and S7j). During the early stages of woody plant expansion under warming and drying conditions, heterotrophic activity increased, as indicated by higher εhopane/water values (Fig. 2i). This coincided with an initial rise and subsequent decline in aCAR (Fig. 2m), suggesting that enhanced heterotrophy offset the gains from increased primary productivity, causing aCAR fluctuations. As woody plants expanded further and peat organic matter composition shifted, elevated aromatic and decreased carbohydrate proportions inhibited heterotrophy, facilitating carbon sequestration (Fig. 3, Pathway 3). Notably, the highest aCARs occurred not in intervals with the maximum woody macrofossil abundance but with the lowest εhopane/water values and a sharp bacterial and fungal biomass decline around 7.1 ka (Fig. 2d, e, i), emphasizing the critical role of heterotrophic metabolism in shaping carbon storage on top of the influence of plant primary productivity10,12,15.

Our study suggests that woody plant expansion initiates a cascade of processes—enhanced primary productivity, shifts toward recalcitrant peat organic matter composition, and suppression of microbial heterotrophy—that can collectively drive substantial increases in carbon accumulation and mitigate the typically adverse effects of climate disturbance such as warming and drying, although likely there are limits to how much warming and drying the system may undergo before the peat soils become aerated. Our data allow us to estimate the increased carbon accumulation driven by woody plant expansion at our site. For every 10% increase in the proportion of woody plants, the combined effects of enhanced primary productivity, changes in peat organic matter composition, and shifts in microbial metabolism contribute to an additional 16.9 g C/m2/yr of carbon accumulation.

However, it is important to recognize that increased woody plant expansion does not necessarily lead to indefinite aCAR increases, as excessive drying can shift peatlands toward aerated conditions, reducing long-term carbon storage. This aligns with growing evidence of a tipping point where further drying no longer enhances carbon accumulation—it has been shown that more dramatic transitions result in greater water table drawdown by evapotranspiration, causing organic carbon oxidation rather than storage60. A similar pattern is observed at ZGT peatland, after 6 ka, woody vegetation declined, herbaceous plants resurged (Fig. 2a), and aCAR decreased (Fig. 2m). This pattern may reflect a response to increasingly dry local conditions and/or possible anthropogenic disturbance, although direct evidence remains lacking.

In addition, during the period of woody plant expansion, woody vegetation co-occurred with abundant Sphagnum (Fig. 2a), indicating that it remained a peatland rather than completing a full transition to forest. This vegetation balance may have played a crucial role in maintaining peat water levels, preventing excessive drying that could otherwise lead to carbon oxidation. This highlights the critical role of Sphagnum in protecting peatland carbon storage through suppressing microbial decomposition61. The synergy between Sphagnum-derived compounds and woody plant-derived aromatics within a symbiotic ecosystem plays a key role in inhibiting microbial metabolism and promoting the preservation of recalcitrant organic matter, ultimately sustaining long-term carbon accumulation.

Additionally, we propose that aCAR should be interpreted with caution, as it represents the current measured value and may not reflect the original accumulation rate due to post-depositional decomposition and autogenic processes. In our ZGT core, lower aCAR and TOC from the Late Pleistocene to the early Holocene (Fig. 2l, m) were likely influenced in part by enhanced decomposition during later warming and drying, as prolonged exposure in an expanded acrotelm delayed incorporation into the catotelm. This underscores the combined influence of initial carbon inputs and secondary decomposition in shaping aCAR. Consequently, the peak aCAR observed during 8–6 ka may indicate a higher original accumulation rate as well as a more recalcitrant organic matter composition that better resisted degradation during subsequent dry periods. This further highlights the pivotal role of vegetation shifts in shaping peat organic matter composition and microbial activity, ultimately regulating long-term carbon storage.

To broaden the understanding of woody plant impacts on peatland carbon accumulation, we compiled apparent carbon accumulation rate (aCAR) and long-term apparent rate of carbon accumulation (LORCA) data from 155 peatlands globally, comprising 196 peat cores and 520 measurements of aCAR/LORCA (Fig. S9; Supplementary Data 1). To minimize autogenic effects, we excluded aCAR/LORCA data from the recent undecomposed surface layers (acrotelm), which can lead to artificially elevated carbon accumulation estimates. We categorized these aCAR/LORCA records based on vegetation dynamics within these peat cores and climate zones, separating northern peatlands (128 sites, 166 cores, 372 measurements) and tropical peatlands (27 sites, 30 cores, 148 measurements). Within each group, we further distinguished between woody-dominated periods (woody plant proportion > 50%) and non-woody periods. In northern peatlands, the mean aCAR/LORCA during woody-dominated periods was 39.9 g C/m2/yr (n = 151), significantly higher than during non-woody periods (28.1 g C/m2/yr, n = 221; p < 0.0001, t = 4.96, df = 370) (Fig. S9a). In tropical peatlands, the mean aCAR/LORCA during woody-dominated periods was even higher, at 83.1 g C/m2/yr (n = 117), compared to 27.8 g C/m2/yr during non-woody periods (n = 31; p = 0.0017, t = 3.202, df = 146) (Fig. S9b). These findings support a consistent role of woody vegetation in enhancing carbon accumulation across climate zones.

While both tropical and northern peatlands show enhanced aCAR/LORCA during woody-dominated periods, we note that the mean aCAR/LORCA in tropical peatlands is substantially higher (Fig. S9b), despite the high rates of microbial respiration expected at higher temperatures56. This likely reflects ecosystem-scale differences in primary productivity, vegetation structure, and peat composition. In particular, tropical swamp forests typically exhibit continuous woody dominance, high litter inputs, and high recalcitrant carbon content51, all of which may act synergistically to enhance long-term carbon accumulation. Therefore, although woody vegetation appears to support greater aCAR/LORCA across peatland types, the underlying mechanisms may differ between tropical and northern systems.

Further comparisons of Holocene carbon accumulation and net carbon pool data from 127 northern peatland sites, alongside woody macrofossil records from 76 sites23,62,63 indicate that continuous woody plant expansion in those settings is also closely linked to increased carbon pools and aCAR (Fig. S9). Notably, accelerated woody expansion in the past 2000 years has likely driven substantial increases in both net carbon pools and aCAR (Figs. S9c, S9d, S9e), although caution should be taken when comparing recent aCAR due to the autogenic effect of the partially undecomposed acrotelm resulting in apparently higher aCAR64,65 that do not necessarily result in higher long-term accumulation.

This more extensive dataset reinforces the idea that woody plant expansion may have a favorable impact on peatland carbon accumulation on a millennial scale, highlighting its importance in maintaining peatland carbon storage amid climatic shifts. Our findings emphasize the crucial interactions between plants, peat organic matter composition, and microbial processes that enable at least some peatlands to sustain long-term carbon storage through the expansion of woody plants. As ecosystems are reshaped by climate change, woody plant expansion offers a promising natural strategy to bolster peatland resilience and enhance carbon sequestration. However, the effectiveness of this feedback will depend on the adaptive responses of microbial communities, stability of hydrological systems, and extent of the woody plant expansion. By integrating past peatland dynamics with future projections, our study underscores the need to understand and harness these mechanisms to maximize the role of peatlands as effective long-term carbon sinks in mitigating global climate change.

Methods

Study site

The ZGT peatland (26°24′40″N, 114°04′40″E) is a mire located in Yanling County, Hunan Province, at an elevation of 1600–1800 m above sea level. It is surrounded by peaks ranging from 1790 to 1830 m and lies at the headwaters of a local stream. The regional vegetation is dominated by evergreen broadleaved forests, while the peatland itself features a mix of shrubs (Camellia caudata, Rhododendron spp.), sedges (Carex spp.), and mosses (Sphagnum palustre, Polytrichum commune) (Fig. 1). The site experiences a mean annual temperature of 14.2 °C and a mean annual precipitation of 1890 mm66.

Three 2-m peat cores were collected in September 2016 using a Russian peat corer in the central area of the ZGT peatland. The cores revealed a surface layer of plant remains, underlain by black humified peat (4–26 cm) and brown peat (26–184 cm). Below 184 cm, the peat transitions to gravelly sediments over granite bedrock35. The peat cores were subsampled at 1 cm intervals in the field, then transported to the laboratory and stored at −20 °C prior to any further processing.

Radiocarbon dating

Radiocarbon dating was performed on 19 selected peat samples, with plant macrofossils such as leaves, seeds, and shoots carefully examined under a microscope prior to analysis. The prepared samples were sent to Beta Analytic Inc. for radiocarbon measurements. The resulting conventional radiocarbon ages were calibrated to calendar years using the Bacon age-depth model. The results indicate that the 184 cm peat core spans the past 14,000 years (Supplementary Fig. S10)35.

Plant macrofossil analysis

Plant macrofossil analysis and identification were conducted following standard protocol20. Briefly, 2 cm3 peat subsamples were treated using a NaOH extraction method. Each subsample was placed in a 100 ml beaker with 5% NaOH solution, heated to boiling, and maintained for 2 hours to remove humic acids that could interfere with microscopic observation. After cooling, the samples were passed through a 125 µm sieve and rinsed thoroughly with distilled water until the filtrate was clear, leaving only plant residues on the sieve. The cleaned residues were transferred with tweezers into labeled centrifuge tubes containing deionized water for further analysis. Plant macrofossil identification was performed using the Quadrat and Leaf Count Macrofossil Analysis (QLCMA) method. The centrifuge tubes were gently shaken, and small amounts of residue were transferred to Petri dishes overlaid with a 10×10 grid. Under a stereo microscope (10× to 50× magnification), residues were examined across 20 randomly selected fields of view. Remaining residues were subsequently analyzed in additional Petri dishes until all materials had been examined. The relative abundance of woody, herbaceous, and moss remains was calculated as percentages of the total macrofossil content.

Lipid extraction and instrumental analysis

Approximately 1.5 g of freeze-dried peat was finely ground to pass through a 60-mesh sieve. The lipids were extracted by sonicating the sample six times for 10 min each using a dichloromethane/methanol mixture (9:1, v/v). An internal standard of androstane, androstanol, and cholanic acid were added prior to extraction. Solvent was removed using a rotary evaporator under reduced pressure, and the resulting extract was fractionated through silica gel column chromatography. Sequential elution with hexane, hexane/dichloromethane (1:1, v/v), and dichloromethane/methanol (1:1, v/v) yielded aliphatic, aromatic, and polar fractions. The aliphatic fraction was analyzed with a gas chromatograph-mass spectrometer (GC-MS; Agilent 6890 GC interfaced with a 5973 mass selective detector) equipped with a DB-5MS column (30 m × 0.25 mm, 0.25 µm film thickness). The GC oven was programmed to start at 70 °C (1 min hold), ramp to 200 °C at 10 °C/min, and then to 300 °C at 3 °C/min (10 min hold). The polar fraction was subjected to saponification to separate it into neutral and acidic sub-fractions. The acidic sub-fraction was then methylated using BF3-methanol to produce fatty acid methyl esters (FAMEs), which were subsequently analyzed using the same instrumentation and protocols as described above.

Compound-specific carbon isotope ratios (δ13C) were determined using a Finnigan Trace GC connected to a Delta Plus XP isotope ratio mass spectrometer with a DB-5MS column (60 m × 0.25 mm, 0.25 µm film thickness). The oven program began at 50 °C (1 min hold), ramped to 220 °C at 10 °C/min (2 min hold), then to 300 °C at 2 °C/min, and finally to 310 °C at 10 °C/min (20 min hold). For δ13C analysis, the oxidation furnace was maintained at 950 °C. Squalane (δ13C = −19.8‰) was used as the internal standard, and system calibration was carried out using n-alkane standards (C16-C30, Indiana University). Carbon isotope reproducibility was better than ±0.5‰. For hydrogen isotope ratios (δ2H), a Finnigan Trace GC coupled to a Delta V Advantage mass spectrometer was used, with a high-temperature conversion furnace at 1400 °C. Squalane (δ2H = −167‰) served as the internal standard, and calibration was verified using n-alkane standards with known δ2H values (Indiana University). Reproducibility for δ2H measurements was better than ±5‰. Peak amplitudes were maintained within 1–4 V for δ2H and 0.5–4 V for δ13C to ensure data quality, with results discarded for peaks below these thresholds. All isotope results were reported in delta notation (‰) relative to VSMOW for δ2H and VPDB for δ13C. To correct for the isotopic contribution of the added methyl group during methylation, we applied a mass balance correction using the following equation:

$${{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{FA}}}}}=[(n+1)\times {{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{FAME}}}}} - {{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{methanol}}}}}]/n$$
(1)

where n is the number of carbon atoms in the fatty acid, δ13CFAME is the measured isotopic composition of the fatty acid methyl ester, and δ13Cmethanol is the δ13C value of the methanol used for methylation. The δ13Cmethanol was determined by methylating phthalic acid of known isotopic composition (reference standard provided by Dr. Arndt Schimmelmann, Indiana University) and calculating the value based on mass balance.

Isotope offset calculation

The isotopic offset for carbon and hydrogen between C31αβ hopane and their respective reference materials was determined using the following equations:

$${\varepsilon }_{{{{\rm{hopane}}}}/{{{\rm{bulk}}}}}=\frac{{{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{hopane}}}}}+1}{{{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{bulk}}}}}+1}-1$$
(2)
$${\varepsilon }_{{{{\rm{hopane}}}}/{{{\rm{water}}}}}=\frac{{{{{\rm{\delta }}}}}^{2}{{{{\rm{H}}}}}_{{{{\rm{hopane}}}}}+1}{{{{{\rm{\delta }}}}}^{2}{{{{\rm{H}}}}}_{{{{\rm{water}}}}}+1}-1$$
(3)

The calculated offsets were scaled by a factor of 1000 and expressed in per mil (‰) following standard reporting conventions48. Here, the δ13Cbulk represents the bulk organic carbon isotopic composition of the peat, and δ2Hwater denotes the isotopic composition of precipitation water. The δ2Hwater values were estimated using the regional meteoric water line and calculated based on the oxygen isotopic composition of carbonate from nearby Dongge Cave, as described by the equation:

$${{{{\rm{\delta }}}}}^{2}{{{{\rm{H}}}}}_{{{{\rm{water}}}}}=7.9\times {{{{\rm{\delta }}}}}^{18}{{{{\rm{O}}}}}_{{{{\rm{carb}}}}}+8.2$$
(4)

Where the δ18Ocarb is the carbonate oxygen isotope value derived from the Dongge Cave67, assumed to represent the regional precipitation oxygen isotope composition. Effects of temperature on precipitation δ18O and on fractionation during carbonate formation were considered negligible, as the temperature dependence of oxygen isotope fractionation have minimal (+0.05‰/°C) impact under southern China’s climatic conditions, and the slope (7.9) and intercept (8.2) correspond to the regional meteoric water line parameters33.

Fourier transform infrared (FTIR) spectroscopy

FTIR spectroscopy was used to analyze the chemical components of peat organic matter51,68. Briefly, the FTIR spectra were obtained using an ALPHA II spectrometer (Bruker, Germany) with KBr pellets prepared by mixing 150 mg of dried KBr with 1 mg of finely ground peat sample. Spectra were acquired over the range of 4000–450 cm−1 at a resolution of 1 cm−1, and baseline corrections were applied to ensure accuracy. A total of 32 scans were averaged for each sample. Key peaks in the spectra were selected to quantify carbohydrates and aromatics (Supplementary Fig. S8), carbohydrates were assessed from the carb peak (1030–1080 cm−1), while aromatics were represented by arom15 (1513–1515 cm−1) and arom16 (1600–1650 cm−1)51. Peak heights were normalized to the total spectral area51, to allow for consistent cross-sample comparisons. The ratios of these normalized peak areas were used to evaluate the peat organic matter composition alterations in the ZGT peat core.

Carbon dynamics and bulk geochemical analysis

The total organic carbon content of the peat samples was determined using a Vario MiCRO cube Element Analyzer. Dry bulk density was measured by determining the mass of peat dried at 50 °C and dividing it by the known sample volume. The apparent carbon accumulation rate (aCAR, g C/m2/yr) was calculated using the following equation69:

$${{{\rm{aCAR}}}}=r/1000\times {\tilde{n}} \times c$$
(5)

where r represents the peat accumulation rate (mm/yr), ñ is the dry bulk density (g/m3), and c is the carbon content (g C/g dry weight).

The δ13Cbulk of peat organic carbon were analyzed using an elemental analyzer coupled to a Finnigan MAT253 isotope ratio mass spectrometer. Prior to analysis, freeze-dried samples were finely ground and sieved through a 200-mesh sieve to ensure uniformity. To remove carbonates, samples were treated with HCl, followed by a secondary freeze-drying step. Approximately 40–60 mg of the treated sample was encapsulated in tin cups for analysis. Calibration of the δ13C values of bulk organic carbon was conducted using certified reference materials of GBW04407 (δ13C = −22.4‰) and GBW04408 (δ13C = −36.9‰). A replicate was analyzed every six samples to monitor consistency, with a reproducibility better than ±0.2‰. A blank correction was applied by analyzing empty tin cups to account for background signals. To ensure consistent peak amplitudes, injection volumes were carefully controlled so that sample signals matched those of the calibration standards.

Global data collection

We compiled carbon accumulation data from 196 peat cores across 155 peatlands worldwide. These data include a total of 520 measurements of apparent carbon accumulation rates (aCAR) and long-term apparent carbon accumulation rates (LORCA). Peatland site information and data sources are provided in the Supplementary Data 1. To avoid overestimation of long-term carbon accumulation, we excluded the significantly elevated aCAR values observed in the uppermost layers of peat cores. These elevated rates are likely influenced by the recent accumulation of undecomposed organic matter that has not undergone secondary decomposition processes. In other words, the autogenic effects of partially undecomposed surface layers (acrotelm) can lead to artificially high aCAR values, but this does not necessarily indicate greater long-term carbon accumulation.

To account for ecological differences between climatic zones, we categorized the compiled aCAR/LORCA records into two groups: northern peatlands and tropical peatlands. Northern peatlands include 128 sites with 166 cores and 372 measurements of aCAR/LORCA, while tropical peatlands include 27 sites with 30 cores and 148 measurements. This separation was made given that tropical swamps are ecologically distinct from northern peatland systems. Within each group, we categorized the carbon accumulation data into woody-dominated and non-woody-dominated periods based on vegetation reconstructions from the same peat cores. A period was classified as woody-dominated when woody plants (including shrubs and trees) accounted for more than 50% of the plant community. For peat cores lacking vegetation reconstructions, classification was based on the overall description of plant remains within the core, peat lithology (e.g., woody peat vs. herbaceous peat), and the modern vegetation composition at the site.

Statistical analyses

All statistical analyses were conducted to assess relationships and differences among variables. For correlation analyses, Pearson correlation coefficients (r) were calculated, and results were reported with significance (p values) and sample size (n). For comparisons between two groups, an unpaired t-test was used to assess significant differences, with results expressed as p-value, t-value, degrees of freedom (df), and sample size (n). For comparisons among three or more groups, such as isotopic differences in n-alkanes among herbaceous plants, woody plants, and Sphagnum mosses (Fig. S2), one-way ANOVA was performed. Results were reported with p-value, F-value, degrees of freedom (df), and sample size (n). The significance threshold was set at p = 0.05. We also applied the coefficient of variation, variance, and subsampling to verify that the increased variability during the 8–6 ka period was not an artifact of a higher number of data points. When sample sizes were equalized, both the coefficient of variation and variance remained higher during 8–6 ka compared to other periods, confirming the robustness of this pattern.