Ecosystem dynamics of an ice-poor permafrost peatland in eastern Eurasia: paleoecological insights into climate sensitivity

Northern peatlands are carbon-rich ecosystems highly sensitive to climate change, with nearly half of their carbon stocks associated with permafrost. Peat-based paleoecological records provide insights into the complex responses of permafrost peatlands to long-term climate variability, but most studies have been conducted in ice-rich permafrost peatlands in Europe and North America. Here, we use multiple active-layer cores to reconstruct the ecosystem history of an ice-poor permafrost peatland in eastern Eurasia, near the southernmost limit of circum-Arctic permafrost but outside the distribution of circum-Arctic thermokarst landscape.

Our results show that the peatland, which has developed on a floodplain since the late Holocene cooling, underwent a major phase of lateral expansion during the Little Ice Age. A fen-to-bog transition occurred in recent decades, with dry-adapted Sphagnum mosses replacing herbaceous vegetation across the site and having rapid surficial peat accumulation. Carbon isotope ratios of Sphagnum macrofossils, a proxy for surface wetness, indicate that Sphagnum mosses were initially established under very dry conditions but that their habitats have since become gradually wetter.

Synthesizing these findings, we highlight that (1) permafrost aggradation during climate cooling may promote new peatland formation over permeable mineral substrate by impeding drainage; (2) anthropogenic climate warming and active-layer deepening can induce an ecosystem-scale regime shift, but ice-poor permafrost peatlands generally exhibit stability and homogeneity due to the absence of dynamic surface morphology (such as frost heave and thermokarst collapse); (3) ongoing wetting may result from surface adjustment–hydrology feedback and vegetation–hydrology feedback, demonstrating the internally driven resilience of ice-poor permafrost peatlands in maintaining their hydrology and carbon accumulation; and (4) ice-poor permafrost peatlands are likely to remain persistent carbon sinks under ongoing and future climate change.

1 Introduction

Northern peatlands are globally important and persistent carbon (C) sinks over time (Gorham, 1991; Yu et al., 2011; Treat et al., 2019), with the size of their C stocks estimated to be 400–500 Pg C (Yu, 2012; Hugelius et al., 2020). Almost half of the northern peatland C is affected by circum-Arctic permafrost (Smith et al., 2007; Hugelius et al., 2020). This biosphere–cryosphere interaction renders permafrost peatlands – commonly found in northern high latitudes – highly sensitive to climate warming and permafrost degradation, which, beyond the simple thawing process, entails compound changes in surface morphology, hydrology, soil structure, vegetation composition, and erosion regime (Payette et al., 2004; O'Donnell et al., 2012; Borge et al., 2017; Olefeldt et al., 2021; Errington et al., 2024). There is a growing concern that degradation of C-rich permafrost peatlands may trigger strong C-cycle feedback that will further accelerate climate warming in the future (Schneider von Deimling et al., 2012; Turetsky et al., 2020).

Permafrost peatlands are heterogeneous landscape features dominated by diverse landforms across permafrost zones (Zoltai and Tarnocai, 1975; Vitt et al., 1994). For example, perennial frost mounds such as palsas or peat plateaus are found in the boreal zone, whereas ice-wedge polygons are found in the Arctic tundra zone (Treat et al., 2016; Olefeldt et al., 2021). These landforms are not always in equilibrium with the climate (Camill and Clark, 1998; Vitt et al., 2000) and may undergo multiple and locally inconsistent stages of transition in response to permafrost degradation (Swindles et al., 2015; Hugelius et al., 2020; Olefeldt et al., 2021), forming new lakes, ponds, fens, and bogs. Generally, field-based greenhouse gas flux measurements and laboratory-based incubation or mesocosm experiments indicate that decomposition of previously frozen organic matter (OM), once upon thaw, has the potential to increase emissions of both CO₂ and CH₄ (and N₂O) (Turetsky et al., 2002; Hodgkins et al., 2014; Voigt et al., 2017, 2019). Direct measurements of belowground C stocks across a chronosequence also present evidence for possibly substantial losses of organic C following thaw and surface collapse (O'Donnell et al., 2012; Jones et al., 2017; Harris et al., 2023). However, these studies yield highly diverse results due to methodological biases, regional variations (permafrost history and peat type), and local processes (geomorphic setting and hydrology) (Treat et al., 2014; Cooper et al., 2017; Estop-Aragonés et al., 2018; Knoblauch et al., 2018; Heffernan et al., 2020; Manies et al., 2021; Heffernan et al., 2024). Therefore, predicting the future fate and trajectory of C stored in permafrost peatlands as well as their climate feedback relies on empirical equilibrium models due to spatiotemporal complexities inherent to such systems (Hugelius et al., 2020; Turetsky et al., 2020; Treat et al., 2021; Fewster et al., 2022).

Paleoecological records derived from peatland archives offer crucial insights into how permafrost peatlands respond to climate variability during their own development histories and are particularly relevant to understanding the slow changes in and interactions among climate, permafrost, hydrology, vegetation, disturbance, and C accumulation on decadal to millennial timescales (Camill et al., 2009; Bauer and Vitt, 2011; Hunt et al., 2013; Jones et al., 2013; Swindles et al., 2015; Gałka et al., 2017; Xia et al., 2024a). Among boreal peatlands, multiple records document that permafrost aggradation during Holocene climate cooling resulted in elevated and drier peatland surfaces, with an increased abundance of bog-like vegetation (shrubs and Sphagnum mosses), more frequent and severe wildfires, and decreased C accumulation rates (Robinson and Moore, 2000; Oksanen et al., 2001; Turetsky et al., 2007; Hunt et al., 2013; Treat et al., 2016). During the current warm period with permafrost thaw, high-resolution proxy records, however, show diverse ongoing ecohydrological trajectories, with some shifting drier due to deeper active layers or higher evapotranspiration rates, some shifting wetter due to accelerated thaw or surface collapse, and some experiencing fluctuating conditions (Zhang et al., 2022). These diverse trajectories have had different impacts on peatland C accumulation rates (Swindles et al., 2015; Zhang et al., 2018a; Sim et al., 2021). Moreover, there is growing evidence of vegetation shift in recently drying permafrost peatlands toward the dominance of Sphagnum mosses (Magnan et al., 2018; Taylor et al., 2019; Chartrand et al., 2023; Piilo et al., 2023; Cleary et al., 2024), a well-known “peat builder” that may enhance the C sink capacity through multiple mechanisms (Clymo and Hayward, 1982; van Breemen, 1995).

Despite the progress to date, it is noteworthy that the current literature and relevant knowledge about the histories and present states of permafrost peatlands are disproportionately built on studies carried out across ice-rich permafrost environments in Europe and North America. Conversely, there are very few counterpart records documented from eastern Eurasia (Treat et al., 2016), which encompasses a vast area and a further southward, inland distribution of circum-Arctic permafrost (Fig. 1a). Therefore, our understanding of the processes and mechanisms governing the response of permafrost peatlands to climate change may be incomplete.

Figure 1 Overview of the study region and site. (a) Circum-Arctic map of permafrost peatland organic carbon (OC) storage from Hugelius et al. (2020), overlain by the thermokarst landscape distribution from Olefeldt et al. (2016) in hatched areas and the location of Tuqiang (TQ) peatland – our study site. (b) Map of Northeast China showing absolute elevation (a.s.l.), the distribution of peatlands or peat-forming wetlands (Xu et al., 2018), the location of Tuqiang peatland, the location of other paleoclimate records later mentioned in this paper (1: Hulun Lake; 2: Yihesariwusu Lake; 3: Huola Basin; 4: Hongtu Peatland; 5: Sifangshan Lake; 6: Tianchi Crater Lake; 7: Tanghongling Mire), the current southern limit of permafrost (Brown et al., 1997), and the inferred southern limits of permafrost during the Holocene Hypsithermal period (HHT) and the Little Ice Age (LIA) in China (redrawn from Jin et al., 2000). GHM: Greater Hinggan Mountains; LHM: Lesser Hinggan Mountains. (c) Google Earth aerial image (© Google Earth 2025, CNES/Airbus) of Tuqiang peatland, showing the locations of cores collected in this study (TQ22) and a previous study (TQ16) by Han et al. (2019). The numbers in parentheses indicate peat initiation ages in cal. yr BP. TQ22-C5 and TQ22-C6 did not reach the basal peat; thus peat initiation ages should be earlier than their oldest ages. (d) Field view of Tuqiang peatland (in front of a hill). (e) Close-up view of Tuqiang peatland vegetation (with whitish snowflakes at the time of field sampling in September 2022). A description of the taxa can be found in Sect. 2.1.

Northern Northeast China is situated at the southernmost limit of circum-Arctic permafrost (Fig. 1b). Regional permafrost degradation caused by climate warming and human activities has been widely documented over recent decades (Jin et al., 2000). Although peatlands are abundant in this region (Xing et al., 2015), only a limited number of sites have been thoroughly described and investigated so far. These studies have mostly focused on greenhouse gas fluxes (Miao et al., 2012a; Yu et al., 2017; Sun et al., 2024), biogeochemical cycling (Wang et al., 2010; Guo et al., 2018; Ramm et al., 2022), and C accumulation patterns (Xing et al., 2015; Liu et al., 2019a) or use peatlands as archives for reconstructing past wildfire, dust, pollution, and upland forest histories (Bao et al., 2012, 2014; Gao et al., 2018; Liu et al., 2025). Only a few studies have aimed to explicitly examine their long-term ecosystem histories using peat-core analysis (Liu et al., 2024b; Xia et al., 2024a). Furthermore, despite the presence of permafrost, this region is not within the modern distribution of circum-Arctic thermokarst landscape (Fig. 1a) due to the underlying ice-poor conditions (Brown et al., 1997; Jin et al., 2016; Olefeldt et al., 2016). Regional peatlands barely show ground expansion- or collapse-related surface terrain characteristics, unlike those ice-rich European and North American permafrost peatlands extensively described in the literature (Olefeldt et al., 2021). Therefore, it remains unclear whether and how climate–permafrost dynamics affect regional peatlands or whether such dynamics differ from other permafrost regions.

Herein, we aim to address these critical data and knowledge gaps by analyzing peat properties, plant macrofossils, and moisture indicators from six radiocarbon-dated active-layer cores to reconstruct the ecosystem history of a boreal ice-poor peatland in permafrost-affected northern Northeast China. Specifically, our objectives are to (1) determine how peatland formation and development, including long-term changes in ecosystem structure and function, responded to past climate variability and (2) understand the role of permafrost in regulating these changes. These objectives are driven by our overarching hypothesis that climate–permafrost interactions and relevant feedback are fundamental in shaping the long-term evolution of peatlands, but due to the absence of dynamic surface morphology (such as frost heave and thermokarst collapse) in ice-poor permafrost environments, the specific processes and mechanisms are expected to differ from those in ice-rich permafrost peatlands. Collecting multiple cores allows us to draw a comprehensive picture of peatland history, capture the spatial heterogeneity of surface features and/or ground properties, and reduce biases that can arise from relying on a single core.

2 Materials and methods

2.1 Study site and sampling

Our study site is Tuqiang peatland (52.94° N, 122.86° E; elevation 475 m a.s.l.) in the northern Greater Hinggan Mountains of Northeast China, near the border with Russia (Fig. 1b). From nearby Mohe weather station (1959–2022), the mean annual temperature is −4.1 °C with above-zero temperatures occurring from May to September. The mean annual precipitation is 443 mm, of which 365 mm falls between May and September. This site is within the “predominantly continuous permafrost” zone based on a national classification, which has been considered more regionally suitable (Ran et al., 2012; Zhang et al., 2021). It differs from the “discontinuous permafrost” zone provided by the International Permafrost Association (Brown et al., 1997) and “sporadic permafrost” zone based on the equilibrium state model (Obu et al., 2019). The permafrost coverage is 65 %–75 %, and the mean active-layer thickness is about 1 m (Ran et al., 2012; Wen et al., 2021). The mean annual ground temperature is about −2 °C at 1.5–2.5 m depth based on the borehole measurements taken nearest to the study site (Li et al., 2022). The soil is dominated by ice-poor frozen soil and ice soil (Fan et al., 2023; Zhang et al., 2023), with ground ice mainly found in the 30 m below the ground surface (Zhang et al., 2024).

Tuqiang peatland developed on a floodplain as part of a larger peatland complex extending along the Fukuqi River (Fig. 1c). Peatland vegetation varies with microform features across the site, with hummocks dominated by Sphagnum mosses (such as nutrient-poor species S. magellanicum and S. capillifolium) and Ericaceae shrubs (Vaccinium uliginosum, Rhododendron tomentosum, and Chamaedaphne calyculata) and hollows dominated by tussock sedges (Eriophorum vaginatum) (Fig. 1e; Miao et al., 2012a; Liu et al., 2024b; Sun et al., 2024). Other notable plant species include Alnus hirsuta, Betula fruticosa, Salix rosmarinifolia, Calamagrostis angustifolia, Carex spp., and Polytrichum spp. The peatland hydrology is mainly controlled by precipitation and evapotranspiration fluxes, as well as seasonally changing thaw depth, without obvious influence from nutrient-rich surface water and groundwater inflows (Liu et al., 2024b; Sun et al., 2024). The maximum thaw depth is 50–70 cm, which occurs in October (Wang et al., 2023; Sun et al., 2024). The water table depth generally varies between 10 and 35 cm below shrub–moss hummocks during the growing season, reaching its peak depth in mid-summer if there is no heavy-precipitation event, based on previous field monitoring data (Miao et al., 2012a; Liu et al., 2024a).

Tuqiang peatland has not developed any outstanding permafrost-related surface terrain features such as palsas, peat plateaus, thermokarst collapses, or ponds (Fig. 1d). As a nationally well known site, it has been studied for over a decade with an eddy covariance tower monitoring greenhouse gas fluxes in the northern section (Yu et al., 2017; Sun et al., 2024).

Table 1 Key information about collected peat cores from Tuqiang peatland.

^a Measured in the field. ^b Decimals are reported to correct core compression or expansion during sample transportation. ^c Gap in Sphagnum peat section is due to changes in macrofossil compositions.

In September 2022, we collected six active-layer cores from Sphagnum-dominated surfaces in Tuqiang peatland using a stainless-steel serrated knife and a 7×7 cm box corer: three from the northern section and three from the southern section (Fig. 1c). Their key information is summarized in Table 1. Briefly, two cores, TQ22-C5 and TQ22-C6, did not reach basal peat and mineral substrate. All cores were stored in the laboratory under frozen conditions.

2.2 Laboratory analysis

The collected peatland cores were cut with a band saw in the laboratory into 1 cm slices, which were subsequently subsampled for all the analyses. Due to complications during transportation from the field to the laboratory, some core sections experienced compression or expansion by a few centimeters compared to the original core lengths measured in the field. Assuming that compression or expansion was uniform across the cores, sample depths were corrected if necessary to match the field-measured core lengths (Table 1).

Physical properties of cores were analyzed at 1 cm intervals following the standard protocol of loss on ignition (Chambers et al., 2010). Briefly, subsamples of 1 cm³ were dried in an oven at 105 °C for 24 h to measure water content and dry bulk density. Then, the dried subsamples were burned in a muffle furnace at 550 °C for 4 h to estimate OM content and ash-free bulk density (AFBD). Bulk peat concentrations of the C and nitrogen (N) of cores were analyzed at 2 cm intervals for peat sections. Additional subsamples were oven-dried, and 1.5–2 mg of the dried material was ground using a ball mill and weighed into tin capsules. Measurements were carried out with an elemental analyzer (EA3000, EuroVector, Italy). The analytical precision (1σ) was about 0.4 % for C and 0.2 % for N concentrations.

All cores were analyzed for ¹⁴C dates at a few selected horizons. They have one date at the depth contiguous to the basal peat or the bottom of the available peat (a few centimeters above for TQ22-C1 due to the general lack of credible macrofossil materials), which we refer to as “basal ages” or “oldest ages”. All cores also have one date close to the depth interpreted as the onset of Sphagnum peat over sedge peat. Sphagnum moss fragments or the aboveground part of vascular plant fragments from 1 cm³ subsamples were separated and concentrated under a stereomicroscope as dating materials and then were freeze-dried. Most dating samples (dry weight of >3 mg) were directly sent to the Keck Carbon Cycle AMS (Keck-CCAMS) Laboratory at the University of California, Irvine (USA), or the Beta Analytic Testing Laboratory (USA) for ¹⁴C measurements. Three dating samples were analyzed at the newly established 0.2 MV MICADAS AMS system housed at Northeast Normal University – the authors' institution (Synal et al., 2007). For these runs, samples were pretreated using the acid–base–acid method and then converted into graphite using an automatic graphitization equipment. Graphitized materials were pressed into cathodes, which were loaded into the MICADAS ion source for ¹⁴C measurements, together with standards and blanks to ensure data accuracy. The analytical precision (1σ) of ${}^{\mathrm{14}}\mathrm{C}{/}^{\mathrm{12}}\mathrm{C}$ was typically better than 2 %.

Plant macrofossils from cores were analyzed at 1 cm intervals for peat sections, except at depths where Sphagnum moss macrofossils were scarce, in which case they were analyzed at 2 cm intervals. Briefly, subsamples of 1 cm³ were immersed in a 2 % NaOH solution for over 6 h, rinsed with distilled water, and then sieved through a 125 µm mesh. The sieved fine debris is unidentifiable organic matter (UOM). This fraction was measured for volume after transfer to a cylinder. The unsieved fraction was examined under a stereomicroscope to identify plant macrofossils. The semi-quantitative Quadrat and Leaf Count Macrofossil Analysis (QLCMA) method was employed (Mauquoy et al., 2010). Average volume percentages of different components of plant remains, including charcoals, were estimated from 20 inspections. Additionally, random individual Sphagnum leaves were identified under a compound microscope to the subgenus level (Sphagnum, Acutifolia, Cuspidata, and Subsecunda), with their respective percentages calculated.

Sphagnum peat sections of cores were analyzed for Sphagnum cellulose carbon and oxygen isotopic compositions, which we use as proxies for local environmental changes, at 1 cm intervals for cores TQ22-C1, TQ22-C2, TQ22-C3, and TQ22-C5 and at 2 cm intervals for cores TQ22-C4 and TQ22-C6. Additional 1–2 cm³ subsamples were used and first treated like macrofossil subsamples. Sphagnum stem tissues (about 20–30 shoots in most cases) were further separated and concentrated after removing their branches under a stereomicroscope. These materials were then extracted for cellulose following the established protocol of the alkaline bleaching method, including treatments with NaClO₂–CH₃COOH and NaOH solutions (Loader et al., 1997; Kaislahti Tillman et al., 2010). Homogenized and freeze-dried cellulose materials were weighed and placed into tin capsules for carbon isotope analysis and silver capsules for oxygen isotope analysis, with 0.1–0.3 mg of material used for each. Measurements were carried out at the Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, using a continuous-flow isotope ratio mass spectrometer (MAT253, Thermo Fisher Scientific, USA) coupled with an elemental analyzer (Flash 2000 HT, Thermo Fisher Scientific, USA). Combustion and pyrolysis modes were set for carbon and oxygen isotope analyses, respectively. By convention, results of isotopic ratios were reported using δ notation (in per mil) referenced to VPDB (Vienna Pee Dee Belemnite) for δ¹³C and to VSMOW (Vienna Standard Mean Ocean Water) for δ¹⁸O. The analytical precision (1σ) was about 0.15 ‰ for δ¹³C and 0.2 ‰ for δ¹⁸O (Xia et al., 2024b).

2.3 Modeling peat accumulation

The results of pre-bomb (before 1950) and post-bomb (after 1950) ¹⁴C dates were calibrated into calendar ages based on IntCal20 and Bomb21 NH1 curves, respectively (Reimer et al., 2020; Hua et al., 2022). We utilized a model of peat mass accumulation based on the proportional decay concept (Clymo, 1984; Clymo et al., 1998) to fit the relationship between near-surface cumulative peat masses and peat ages by combining all ¹⁴C-dated horizons with post-bomb ages from six cores. This practice assumes that near-surface peat accumulations in different cores are controlled by very similar rates of mass input (litter production) and output (rapid aerobic decay) and that peat of post-bomb ages is still in the acrotelm (surface “aerobic layer”) and has not yet been transferred to the catotelm (deep “anaerobic layer”). Considering this model allowed us to assess the post-bomb chronology and reasonably reject unrealistic calibrated ages (calibration of post-bomb ages would yield a possible age from before 1963 and another possible age from after 1963). Additionally, this model yielded key parameters fundamental to understanding the control of peatland C balance.

The model assumes that for a given mass of fresh OM m₀ (g cm⁻²) produced annually per unit area, the decay process per unit time follows

$$\begin{array}{}\text{(1)}& {\displaystyle \frac{\mathrm{d}m}{\mathrm{d}t}}=-m\mathit{\alpha }{\left({\displaystyle \frac{m}{m_{\mathrm{0}}}}\right)}^x,\end{array}$$

where m is mass per unit area (g cm⁻²), t is time (years), α is the decay coefficient (yr⁻¹), and x is an exponent controlling the recalcitrance effect (Morris et al., 2015). As decay progresses, the actual decay rate, $\mathit{\alpha }(m/m_{\mathrm{0}}{)}^x$, may either remain constant (x=0) or decrease (x>0) over time, particularly more rapidly with higher x. For a peat column where fresh OM is continually added from the top and decayed OM remains and accumulates over time, the cumulative peat mass (M) can be expressed as the time integral of m:

$$\begin{array}{}\text{(2)}& M\left(t\right)=\underset{\mathrm{0}}{\overset{t}{\int }}m\mathrm{d}t=\left\{\begin{array}{ll}\frac{m_{\mathrm{0}}\left(\mathrm{1}-e^{-\mathit{\alpha }t}\right)}{\mathit{\alpha }}& \text{if }x=\mathrm{0},\\ \frac{m_{\mathrm{0}}\left[{\left(\mathrm{1}+x\mathit{\alpha }t\right)}^{\frac{x-\mathrm{1}}{x}}-\mathrm{1}\right]}{\left(x-\mathrm{1}\right)\mathit{\alpha }}& \text{if }x>\mathrm{0}\text{ and }x\ne \mathrm{1},\\ \frac{m_{\mathrm{0}}\ln \left(\mathrm{1}+\mathit{\alpha }t\right)}{\mathit{\alpha }}& \text{if }x=\mathrm{1}.\end{array}\right.\end{array}$$

Equation (2) is a more general expression of the mass–age relationship compared to several special cases shown in Clymo et al. (1998). It was utilized to fit the cumulative peat mass and age data using the least-squares method, resulting in a potential range of values for m₀, α, and x when combined. The fitting was carried out using an iterative orthogonal sampling technique that takes m₀ and α values from a space of 0–0.2 g cm⁻² and 0–0.2 yr⁻¹ after assigning a particular x value (0–10). The best fit under each x value was accepted only if the coefficient of determination (R²) was > 0.95. A similar fitting practice has been shown by Charman et al. (2013).

In order to interpret peat-core proxy data as time series, we developed age–depth models for the collected cores using piecewise linear interpolations between dated intervals (Zhang et al., 2018a; Sim et al., 2019; Fewster et al., 2023; Cleary et al., 2024). The way we derived age–depth models is based on the fact that our cores have very rapid accumulation rates of near-surface peat but much slower accumulation rates of deeper peat that cannot be dated in high resolution. This approach contrasts Bayesian modeling, which needs to incorporate smoothness-related prior information.

2.4 Other data analyses

To summarize the pattern of peatland vegetation changes, we performed principal component analysis (PCA) to reduce the dimensionality of the macrofossil data, in which important macrofossil types (mean percentage >1 %) were selected and percentage data were log-transformed.

To quantify changes in peatland C capacity, we calculated long-term apparent C accumulation rates (aCARs; alternatively, “LORCA”) for entire peat-core columns, where the C content for OM is based on the mean value (49.7 %) derived from our specific C concentration measurements. These aCAR values represent the net rates of C sequestration during the peat accumulation history and were compared with available net ecosystem C balance (NECB) data derived from contemporary eddy covariance measurements combined with other observations (Sun et al., 2024). We did not present depth-specific aCAR data as time series due to the well-recognized artifact of an increasing trend from old to young peat (Young et al., 2019, 2021).

Figure 2 Lithology information and peat properties (organic matter content, ash-free bulk density, $\mathrm{C}/\mathrm{N}$ ratio) of six cores. Arrows indicate the ¹⁴C-dated horizons.

3 Results

3.1 Peat properties

Peat property data are presented in Fig. 2. Among all subsamples of six cores, the OM content varies from 2.1 % to 100 % and the AFBD varies from 0.014 to 0.311 g cm⁻³. The OM content profiles show lower values at increasing depth and clearly distinguish the transition from mineral substrate to peat in four cores, which is defined based on the threshold of 30 % OM content (Table 1). The AFBD profiles generally show higher values in sedge peat sections than in young Sphagnum peat and old mineral substrate sections (macrofossil data are described in a following subsection). The mean OM content is 92.2 %, 73.4 %, and 7.9 % and the mean AFBD is 0.044, 0.150, and 0.090 g cm⁻³ for Sphagnum peat, sedge peat, and mineral substrate, respectively. Among these values, it is noted that the Sphagnum peat AFBD is very low and the sedge peat AFBD is very high at our site, compared to the northern peatland database (Loisel et al., 2014).

Among all subsamples of six cores, the $\mathrm{C}/\mathrm{N}$ ratio (mass ratio) of bulk peat varies from 16 to 77. The $\mathrm{C}/\mathrm{N}$ ratio profiles show maximum values at shallow depths and much lower values in deeper sections. The stratigraphical transitions into very low $\mathrm{C}/\mathrm{N}$ ratios with greater depths are related to abrupt changes in botanical composition from Sphagnum peat to sedge peat in each core. The mean $\mathrm{C}/\mathrm{N}$ ratio is 49 for Sphagnum peat and 23 for sedge peat. Both values are low compared to the northern peatland database (Loisel et al., 2014).

Table 2 Summary of radiocarbon data from Tuqiang peatland.

^a Lab number in such a format means that the sample was analyzed at Northeast Normal University, China. ^b Lab number with “UCIAMS”: analyzed at the Keck-CCAMS Laboratory at the University of California, Irvine, USA. ^c Lab number with “Beta”: analyzed at the Beta Analytic Testing Laboratory, USA. ^d Rejected outlier date. ^e Selected age based on the stratigraphic order or the mass–age relationship.

3.2 Radiocarbon dating

Radiocarbon dates and other relevant information are listed in Table 2. We obtained many post-bomb dates, even at >40 cm depth, pointing to very rapid accumulation rates of near-surface peat. The basal ages in median values are 610, 660, 75, and 600 cal. yr BP (equivalent to 1340, 1290, 1875, and 1350 CE) from cores TQ22-C1, TQ22-C2, TQ22-C3, and TQ22-C4, respectively (Fig. 1c). The bottom of available peat in core TQ22-C6 has the oldest date, at a median age of 2850 cal. yr BP.

Figure 3 Results of near-surface peat accumulation modeling. (a) The relationship between cumulative peat mass (M) and time (t) derived by combining all post-bomb dates of six cores. Rejected ages after post-bomb calibration are shown as open circles. Accepted ages (closed circles) were fitted using the general expression of Clymo's model, with fitted curves plotted in different colors for different assumptions of recalcitrant effects (x). (b) The fitted values of annual production m₀ and decay coefficient α under different x values.

3.3 Modeled peat accumulation

Figure 3a shows the biplot between cumulative peat masses and calibrated ages of post-bomb dates. By considering the stratigraphic order and the expected concave pattern of the mass–age relationship, we determined, for each post-bomb date, which of the two calibrated ages should be rejected, and then we can accept the other one as the most likely age for that dating horizon. However, we cannot definitively determine the correct ages for two post-bomb dates: 20–21 cm of TQ22-C1 and 34–35 cm of TQ22-C6. For these dates, calibration yielded two ages that are close to each other, making it difficult to determine which is more reasonable (Fig. 3a and Table 2). We accepted their older calibrated ages due to them being closer to other dates with similar cumulative peat masses.

Combining all these accepted calibrated post-bomb ages, we fitted the mass–age relationship using the general expression of peat decomposition model (Eq. 2) and found that the fitted curves can be robust for a wide range of recalcitrance values up to x=1.6 (Fig. 3a), in which stronger recalcitrance requires higher rates of production and decomposition (Fig. 3b). The choice of x=0.8 (assuming an intermediate recalcitrant effect) produced the highest fit (R²=0.962), with m₀=0.087 g cm⁻² and α=0.094 yr⁻¹. Assuming no recalcitrant effect would yield m₀=0.07 g cm⁻² and α=0.039 yr⁻¹ (R²=0.959), both of which are higher than reported values from the acrotelm of other northern peatlands (Yu et al., 2001; Loisel and Yu, 2013).

3.4 Plant macrofossil records of peatland vegetation changes

Full plant macrofossil diagrams are presented in Figs. A1–A6. Macrofossil analysis reveals transitions from lower sedge (Cyperaceae) peat to upper Sphagnum peat in all cores, although cores TQ22-C1 and TQ22-C6 exhibit somewhat fluctuating abundances of Cyperaceae and Sphagnum macrofossils across the transition. We define the onset of Sphagnum peat over sedge peat in each core as the greatest depth where Sphagnum macrofossils exceed 40 % abundance (Table 1 and Fig. 2). The chronology of this stratigraphical transition is well constrained in all six cores. The median ages are 1995, 1955, 1958, 1957, 2001, and 1838 CE, respectively.

Figure 4 Results of principal component analysis of the macrofossil dataset. (a) Component loadings of variables and scores of samples. (b) Changes in PC1 score from different cores with a close-up view of the last 70 years. The black line is the 20-year mean of all individual cores. The onsets of Sphagnum peat are marked on the x axis.

PCA identifies the statistical relationship among different macrofossil types (Fig. 4a). The first component explains 31.1 % of variance and separates the contrasting environmental preferences between dry-adapted Sphagnum mosses (subgenus Sphagnum and Acutifolia) and wet-favoring herbaceous taxa (Cyperaceae). However, charcoal fragments, which are often interpreted as a drought indicator, are tightly associated with herbaceous taxa. The second component explains 12.1 % of variance and separates non-Sphagnum mosses and Ericaceae, likely related to microhabitat conditions or competition for resources. The PC1 score shows a trend of gradual increase since several hundred years ago, followed by an abrupt increase during the recent century related to the onsets of Sphagnum peat accumulation (Fig. 4b).

Figure 5 The C sink capacity of Tuqiang peatland. Long-term apparent C accumulation rates (aCARs) of cores collected in this study (TQ22) and a previous study (TQ16) by Han et al. (2019) are plotted against the ages of their basal peat or the bottom of available peat. The annual net ecosystem C balance (NECB) values for the years 2016 and 2017 (Guo et al., 2018; Sun et al., 2024) are plotted on the left side for comparison.

3.5 Apparent peat carbon accumulation rates

The long-term aCARs for six cores show an age-dependent pattern with the highest aCAR (163.8 g C m⁻² yr⁻¹) found in the youngest core, TQ22-C3, and the lowest aCAR (10.4 g C m⁻² yr⁻¹) found in the oldest core, TQ22-C6 (Fig. 5). The other four cores are of similar ages (660–500 cal. yr BP) but still have highly variable aCARs (24.3–51.2 g C m⁻² yr⁻¹), showing within-site heterogeneity in the rate of C sequestration. In comparison, the currently available NECB at the same site (Fig. 5), derived mostly from atmospheric C flux measurements, indicates a further higher rate of C sequestration at about 55 and 74.2 g C m⁻² yr⁻¹ in 2016 and 2017, respectively (Sun et al., 2024), after additionally accounting for the small fluvial C losses of 4.7 g C m⁻² yr⁻¹ (Guo et al., 2018).

3.6 Isotopic records of environmental changes from Sphagnum peat sections

Among all cores, Sphagnum cellulose δ¹³C (δ¹³C_cell) values range from −32.9 ‰ to −27.9 ‰ and cellulose δ¹⁸O (δ¹⁸O_cell) values range from 14.1 ‰ to 20.2 ‰. We applied a correction for the “Suess effect” using the method by Leuenberger (2007), which removed the effect of changing δ¹³C values of atmospheric CO₂ due to fossil fuel combustion in recent centuries. The “Suess effect”-corrected δ¹³C_cell values range from −31.7 ‰ to −25.9 ‰. Hereafter, δ¹³C_cell data mentioned in this paper all refer to their corrected values.

Figure 6 Isotopic records of environmental changes from the period 1960–2022 from high-resolution near-surface peat of six cores. (a) Annual temperature and precipitation records from Mohe weather station, located about 24 km from the Tuqiang peatland. (b) “Suess effect”-corrected Sphagnum δ¹³C_cell records plotted separately for three cores showing progressively wetter conditions (higher δ¹³C_cell) and three other cores that do not show clear trends. Data from earlier periods (before 1960) are plotted along the left y axis. (c) Similar to (b) but for Sphagnum δ¹⁸O_cell records. (d) Amount-weighted mean growing-season (May–September) precipitation δ¹⁸O for the location of Tuqiang peatland simulated from two nudged isotope-enabled general circulation models, LMDZ4 (Risi et al., 2010) and IsoGSM2 (Yoshimura et al., 2008), the outputs of which are available from the period 1979–2021.

The δ¹³C_cell records show increasing trends with weak fluctuations since Sphagnum mosses became the dominant vegetation in three cores: TQ22-C2, TQ22-C3, and TQ22-C5 (Fig. 6b). However, their increasing trends are not synchronous among different cores. Specifically, the δ¹³C_cell values of TQ22-C3 increase during the period 1960–1990 and remain stable after that, whereas the δ¹³C_cell values of TQ22-C2 increase during the entire period of 1960–2020. The δ¹³C_cell values of TQ22-C5 increase after the 2000s. In the other three cores, the δ¹³C_cell records do not have clear trends but show weak signs of slightly increasing trends in the cores TQ22-C1 during the recent decade and TQ22-C6 since the 1970s (Fig. 6b).

The δ¹⁸O_cell records are characterized by large sample-to-sample fluctuations with no clear trends in all cores (Fig. 6c). This temporal pattern is entirely different from that of the δ¹³C_cell records. However, chronological uncertainties arising from overly simplistic age–depth models preclude determining whether those δ¹⁸O_cell spikes are temporally coherent among different cores or correlating δ¹⁸O_cell variations with the inter-annual variability in temperature, precipitation, or precipitation δ¹⁸O (Fig. 6a, d).

4 Discussion

4.1 Possible climate-driven peatland initiation and lateral expansion

A previous global data synthesis of peatland basal ages revealed that growing-season warming during the last deglaciation and early Holocene is the primary driver of the initiation of the most northern peatland areas by enhancing plant productivity (Morris et al., 2018). However, that study included only a limited number of sites from Northeast China. For our site, Han et al. (2019) obtained a basal age of 2620 cal. yr BP (interpolated from their dates between mineral materials and peat) from its northern section, based on one peat core (TQ16) collected from a location west of ours (Fig. 1c). This is much older than our new basal ages and can be interpreted as the timing of peatland initiation. In the southern section, the oldest age of peat cores collected in this study is 2850 cal. yr BP, but it is not the age of basal peat as we did not reach mineral substrate at this coring location (Fig. 1c). Obviously, basal age data from a single site are insufficient for establishing links with large-scale climate forcing. A previous region-wide and comprehensive data synthesis of Northeast China has revealed that the timing of peatland initiation is spatially variable, with more frequent earlier initiations in warmer, lower-latitude locations (Xing et al., 2015). In northern Northeast China including the Greater Hinggan Mountains relevant to this study and the nearby Lesser Hinggan Mountain region, peatland initiation mainly occurred during the late Holocene, i.e., from 4200 cal. yr BP to the present (Fig. 7a; Xing et al., 2015).

Figure 7 Regional peatland initiation and climate histories during the Holocene. (a) Frequency of basal peat ages from the permafrost-affected Greater Hinggan Mountain and Lesser Hinggan Mountain regions, previously compiled by Xing et al. (2015) and Zhang et al. (2020b). Ages of initiation and lateral expansion of Tuqiang peatland are shown as open triangles and dots. (b) Temperature changes reconstructed using pollen-based transfer function from Hulun Lake (Wen et al., 2010) and using alkenone biomarkers from Yihesariwusu Lake (Jiang et al., 2024). (c) Precipitation changes from a synthesis of multiple proxy records (see Fig. 1b) in northern Northeast China by Zhang et al. (2025), including sites at Huola Basin (Li et al., 2019), Sifangshan Lake (Liu et al., 2017), Tianchi Crater Lake (Zhou et al., 2016; Liu et al., 2019b), and Tanghongling Mire (Yang, 2003), as well as from the model simulation of TraCE-21ka for the same region (Zhang et al., 2025).

The Holocene climate variability has been a subject of debate, particularly regarding (1) whether the Holocene is characterized by long-term cooling as documented by proxy records or long-term warming as simulated by climate models (the Holocene temperature conundrum; Liu et al., 2014b) and (2) whether the early Holocene was wetter than the late Holocene across all monsoon-affected Asian regions as shown in Asian speleothem δ¹⁸O records of monsoon intensity (Wang et al., 2005; Liu et al., 2014a). Fortunately, new studies based on quantitative reconstructions and model–data comparisons provided a timely constraint, showing that the late Holocene – a period of frequent peatland initiations – was characterized by colder and wetter conditions in northern Northeast China compared to the early Holocene (Fig. 7b, c; Jiang et al., 2024; Zhang et al., 2025). This is in remarkable contrast to the global pattern of warming-driven peatland initiations (Yu et al., 2010; Morris et al., 2018).

Previous data syntheses and many site-specific studies showed that after peatland initiation, the peak period of subsequent lateral expansion in northern peatlands occurred during late Holocene cooling (Korhola et al., 2010; Weckström et al., 2010; Ruppel et al., 2013; Quik et al., 2023). These studies suggest that peatland lateral expansion might also be climate-driven, with colder and wetter conditions during these periods facilitating the paludification of surrounding mineral soils (Ruppel et al., 2013). For our site, we obtained three (out of six) peat cores that have young basal peat of similar ages (660–500 cal. yr BP), interpreted as indicating a phase of major peatland lateral expansion. The direction of expansion appears to be from upland toward the river channel (Fig. 1c). The timing is within the Ming–Qing cold period in China (1321–1920), which can be seen as the regional expression of the Little Ice Age (LIA) (Ge et al., 2013; Neukom et al., 2019). A new study using multi-proxy data further suggested that the LIA was characterized by wetter conditions in northern Northeast China (Xia et al., 2024b). Although the lateral expansion of peatlands is poorly documented in our study region, the finding that it occurred at Tuqiang peatland during a climate window of colder and wetter conditions is consistent with the previously identified general pattern in northern peatlands (Ruppel et al., 2013).

Net peat accumulation results from a positive balance between the plant productivity and OM decomposition (Yu et al., 2009). Why are colder and wetter conditions favorable for forming new peatland areas regionally? Superficially, this would suggest that better OM preservation – typically expected under such conditions by slowing down microbial activity – plays a prominent role in initializing and maintaining net peat accumulation. If true, how do we reconcile this finding from the literature that highlights the stronger role of temperature-controlled productivity in shifting the OM balance (Yu et al., 2009; Gallego-Sala et al., 2018; Morris et al., 2018)?

A distinct setting of our study region is that peatlands are primarily found on river floodplains and underlain by permafrost. Early surveys noted a spatial overlap between peat-forming wetlands and permafrost in Northeast China, interpreted as evidence for a symbiotic relationship (Sun, 2000). Geological studies mapping past permafrost boundaries showed that Northeast China was nearly permafrost-free during the warmest period of the Holocene, but the permafrost boundary moved southward by approximately 300 km during the late Holocene and the LIA (Fig. 1b; Jin et al., 2000, 2016). This permafrost history provides a framework for formulating our hypothesis that the transition from seasonally frozen ground to permafrost might be an agent in regional peatland formation. We propose that permafrost aggradation and its possible expansion toward rivers during cooling periods at our site resulted in thinning active layers and reduced water infiltration. These changes likely triggered feedback mechanisms that enhance poorly drained, waterlogged soil conditions locally (Sun, 2000; Ishikawa et al., 2005; O'Donnell et al., 2012), leading to further better OM preservation than by climate cooling and wetting alone and eventually net peat accumulation (Sun, 2000; Treat et al., 2014).

This hypothesis is based on the fact that Tuqiang peatland developed in a floodplain environment, with gravel-/sand-rich sediments observed at the base of cores (Fig. 2). Without permafrost as a physical barrier, rapid and deep drainage through such coarse-grained, highly permeable substrate would prevent soil saturation and also enhance export of dissolved organic C via subsurface flow in a hydrologically connected catchment (Frey and Smith, 2005; Olefeldt and Roulet, 2014; Debolskiy et al., 2021).

Some studies conducted in higher-latitude peatlands have noted that permafrost aggradation may result in the assimilation of undecomposed material into frozen ground, leading to the upward accumulation of labile peat concurrent with syngenetic permafrost (O'Donnell et al., 2012; Treat et al., 2016). In this regime, peat may contain cryostructures (O'Donnell et al., 2012) and often has a very high $\mathrm{C}/\mathrm{N}$ ratio due to limited C mineralization (Sannel and Kuhry, 2009; Treat et al., 2016; Haugk et al., 2022). However, neither was observed at our site. The $\mathrm{C}/\mathrm{N}$ ratio of basal and lower peat sections is only about 25 (Fig. 2), showing no sign of freeze protection, compared to >60 in typical permafrost-affected old peat (Treat et al., 2016). The botanical composition of the peat is dominated by sedges, indicating a prolonged status of wet minerotrophic fen in early peatland history (Figs. 2 and A1–A6). These results provide support for the claim that permafrost aggradation contributed new peat accumulation at our site primarily by maintaining suitable hydrological conditions, rather than by directly freezing the OM.

Together, our analysis links the distinct climate drivers of peatland initiation and lateral expansion with permafrost–hydrology feedback, underscoring the importance of considering surficial geology and cryosphere dynamics in understanding the diverse pathways of peatland formation.

4.2 Vegetation and carbon accumulation dynamics

Plant macrofossils preserved in peatland archives have been used as indicators for long-term ecosystem structure shifts (Hughes and Barber, 2004; Camill et al., 2009; Bauer and Vitt, 2011; Treat et al., 2016). For our site, the macrofossil data show the persistent stability of a wet minerotrophic fen environment dominated by herbaceous vegetation since >2000 years ago, until recent decades when all cores document a consistent, abrupt transition in vegetation composition toward a present-day dominance of dry-adapted Sphagnum mosses (S. subg. Sphagnum and S. subg. Acutifolia) and an increased abundance of shrubs as well as dry-adapted Polytrichum spp. and Aulacomnium palustre (Faber et al., 2016), all indicating a dry ombrotrophic (rain-fed) bog environment (Figs. 2 and A1–A6). This transition in core TQ22-C6 appears slightly more complex: Sphagnum mosses were present somewhat earlier (median age 1838 CE) and had undergone both abundance fluctuations and species turnovers, but the vegetation composition similar to that in other cores was eventually established in recent decades (Fig. A6). For newly formed peat at core TQ22-C3, a fast sedge–Polytrichum/Aulacomnium–Sphagnum succession can be distinguished (Fig. A3). Principal component analysis of the full macrofossil dataset further reveals that the recent abrupt transition to bog-like vegetation had been preceded by a longer-term slow shift in the same direction since the LIA (Fig. 4b).

While a previous study at the same site has already documented this abrupt transition in vegetation composition, which specifically occurred at about 1895 CE based on the analysis of a single peat core (TQ16; Fig. 1c) (Liu et al., 2024b), our new data from multiple cores make a strong case that this is a site-wide phenomenon of fen–bog transition (FBT) with clear signs of regime-shift behavior, i.e., gradual changes followed by sudden shifts (Scheffer and Carpenter, 2003; Loisel and Yu, 2013; Loisel and Bunsen, 2020).

Instrumental climate data from northern Northeast China indicate a temperature increase by up to 4° C since the end of the nineteenth century (Zhang et al., 2011; Yao et al., 2021). The recent FBT at our site underscores the high sensitivity of ecosystem structures in such young, shallow, and permafrost-underlain peatlands to rapid and large-magnitude climate warming. Loisel and Yu (2013) previously described a conceptual model of “warming-driven FBT” for non-permafrost peatlands to explain how anthropogenic climate warming facilitates Sphagnum moss establishment and oligotrophication through surface drying and ecohydrological feedback. For permafrost peatlands, however, we anticipate that the additional permafrost–hydrology feedback further reinforces the processes underlying this model: deepening active layers – the first-stage response to warming-driven permafrost degradation (Swindles et al., 2015; Hugelius et al., 2020) – would further draw down water tables to enhance surface drying. This is supported by our Sphagnum δ¹³C_cell data, which show very low values (corrected values less than −30 ‰) when Sphagnum mosses initially replaced herbaceous vegetation in macrofossil records (Fig. 6b), indicating extremely dry conditions when FBT occurred (see the following section, Sect. 4.3, on interpretations of Sphagnum δ¹³C_cell data). Several other studies have documented recent FBT or Sphagnum moss expansion in permafrost peatlands elsewhere (Magnan et al., 2018; Taylor et al., 2019; Magnan et al., 2022; Chartrand et al., 2023; Piilo et al., 2023), likely driven by mechanisms similar to those in our record.

There is extensive literature describing the highly dynamic and heterogenous nature of permafrost peatlands with complex ecosystem shifts in response to natural climate variability (Zoltai, 1993; Jones et al., 2013; Gao and Couwenberg, 2015; Treat et al., 2016; Zhang et al., 2018b; Sim et al., 2021; Piilo et al., 2023; Xia et al., 2024a). Specifically, climate-driven permafrost aggradation and degradation can strongly alter surface morphology, forming palsas, peat plateaus, thermokarst collapses, fen–bog complexes, and lakes/ponds, with each following different succession pathways (Olefeldt et al., 2021). However, Tuqiang peatland is a site showing historical stability, along with homogeneity in recent ecosystem shifts and a present-day structure free of relict permafrost landforms (Fig. 1d). We attribute such major differences to the low ice content of regional permafrost compared to the results of previous studies conducted in Europe and North America (Brown et al., 1997; Jin et al., 2016; Olefeldt et al., 2016). Again, our study region is outside the distribution of circum-Arctic thermokarst landscape due to ice-poor conditions (Olefeldt et al., 2016, 2021).

The climate warming, active-layer deepening, and, in particular, the recent transition to the dominance of Sphagnum mosses may collectively alter the balance between plant productivity and OM decomposition, with profound consequences for the peatland C sink function (Clymo and Hayward, 1982; van Breemen, 1995; Loisel and Yu, 2013). Some studies have attempted to interpret time series of depth-specific aCARs to understand the temporal shifts in the rate of the C sink (Swindles et al., 2015; Zhang et al., 2018a, 2020a; Sim et al., 2021; Cleary et al., 2024), but direct comparisons between recent and long-term rates can be questioned (Young et al., 2019). Here, we found that the present-day NECB, estimated over 2 monitored years (Guo et al., 2018; Sun et al., 2024), is much higher than the centuries- and millennia-averaged NECB, estimated from long-term aCARs across entire peat-core columns (Fig. 5). This suggests a potential stronger C sink capacity under present conditions compared to the past. However, we caution that the present-day NECB is likely an overestimate due to the short measurement period not accounting for low-frequency disturbances (e.g., wildfire) and the inconsistent methodology (C fluxes are estimated using the concentration gradient method in freezing winter) (Miao et al., 2012b; Sun et al., 2024).

At our site, Sphagnum peat accumulation is quite rapid, reaching a thickness of about 40 cm of material over several decades (Fig. 2). Peat accumulation modeling quantifies near-surface peat buildup as being governed by a stable and rapid input of plant litter and stable and rapid proportional decay, with possible recalcitrance in recently formed OM (Fig. 3). Therefore, the present-day stronger C sink is likely related to the persistent surficial peat accumulation from Sphagnum litter, which must greatly outweigh any deep C loss caused by deeper active layers and aerobic conditions. Additionally, the generally low $\mathrm{C}/\mathrm{N}$ ratios of basal and lower peat sections indicate the lack of freeze protection and poor decomposability (Treat et al., 2016; Haugk et al., 2022). Together, our analysis suggests that anthropogenic warming-driven FBT is likely amplified by permafrost–hydrology feedback, with no sign of a reduction in the C sink. In order to understand possible mechanisms for the persistence of the C sink function, we next investigate the pattern of recent environmental changes documented in isotopic ratios of Sphagnum macrofossils.

4.3 Ongoing trend toward wetter conditions after the fen–bog transition

Isotopic records from Sphagnum macrofossils offer further insights into changes in surface conditions since Sphagnum mosses became dominant at our site, providing essential context for understanding ongoing environmental trajectories and their drivers. Sphagnum δ¹³C_cell is a proxy for surface wetness based on the “water film” mechanism, which has been well established through multiple studies (Rice and Giles, 1996; Williams and Flanagan, 1996; Loisel et al., 2009; Xia et al., 2020, 2025). Briefly, higher δ¹³C_cell values indicate wetter conditions, as a thicker water film on leaf surfaces reduces isotopic fractionation caused by carboxylation (Farquhar et al., 1989). For peatlands like ours relying on precipitation as the primary source of water, Sphagnum δ¹⁸O_cell is a proxy for precipitation δ¹⁸O (Daley et al., 2010; Granath et al., 2018; Xia et al., 2018), which is controlled by multiple atmospheric drivers such as temperature, precipitation amount or seasonality, and moisture sources (Dansgaard, 1964).

Although the recent FBT phenomenon means that the peatland has become much drier than it used to be, the δ¹³C_cell records indicate that the peatland has ongoing gradual trends, lacking fluctuations and being asynchronous among different cores, from very dry to wetter conditions after the FBT (Fig. 6b). Such a pattern of moisture changes cannot be simply attributed to a common driver of precipitation, which exhibits large inter-annual variability (Fig. 6a). By contrast, the δ¹⁸O_cell records show no clear trend but instead contain large sample-to-sample fluctuations (Fig. 6c), coinciding with the pronounced inter-annual atmospheric variability that may affect precipitation δ¹⁸O (Fig. 6a, d). Therefore, our Sphagnum δ¹³C_cell and δ¹⁸O_cell records present two seemingly conflicting phenomena: evidence for the rain-fed nature of Sphagnum moss habitats yet no direct effects on their wetness. Here, we interpret these data as indicating that Sphagnum wetness is controlled by internal, locally paced surface processes. As described below, we specifically hypothesize that such internal dynamics can be attributed to gradual subsidence from the degradation of ice-poor permafrost or the ecohydrological feedback driven by Sphagnum mosses acting as ecosystem engineers.

Previous studies have proposed an empirical stage model predicting that active-layer deepening and surface drying in permafrost peatlands would be succeeded by ground collapse, thermokarst formation, and land surface wetting once thaw progresses into ice-rich layers (Swindles et al., 2015; Hugelius et al., 2020). A newly published global synthesis of peatland water table reconstructions also revealed that those permafrost peatlands showing an ongoing wetting trend had experienced drier “pre-conditions” before the wet shifts (Sim et al., 2021; Zhang et al., 2022). Therefore, the ongoing trend from very dry to wetter conditions over the last several decades documented in our records appears to be an expected outcome for peatlands under warming-driven permafrost degradation. However, field observations at Tuqiang peatland show no sign of peat shrinkage, cracking, or collapse – features that characterize the wetting stage (Fig. 1d). Therefore, unlike the empirical stage model developed for ice-rich peatlands in Europe and North America, we hypothesize that the wetting may be linked to surface subsidence resulting from a combination of gradual thaw and decomposition-induced peat loss, which raised the water table when the effect of active-layer deepening slowed down. For simplicity, we term this general mechanism as surface adjustment–hydrology feedback (Waddington et al., 2015).

Sphagnum mosses are well-known ecosystem engineers that possess multiple plant properties to modify local environmental conditions in ways that favor their success and peat formation (Clymo and Hayward, 1982; van Breemen, 1995; Rydin and Jeglum, 2013). Among these properties, they have an extraordinary water-holding capacity, typically containing several to over 10 times more water than their tissue weight within their hyaline cells and external pore spaces, and they can maintain a capillary fringe to distribute water under various conditions (Clymo and Hayward, 1982; Rydin and Jeglum, 2013). Additionally, Sphagnum peat is a poor heat conductor, and as an insulating layer, it may further contribute to slowing down active-layer deepening (van Breemen, 1995; Park et al., 2018). Therefore, we hypothesize that the wetting may also have resulted from plant-mediated ecohydrological feedback (vegetation–hydrology feedback) regulating local moisture dynamics as Sphagnum peat gradually built up.

Overall, the isotopic records document internally driven, progressively wetter conditions at Tuqiang peatland over the last several decades, which occurred against a background of long-term drying that had led to the FBT. These two phenomena indicate that anthropogenic climate warming and concurrent permafrost degradation would not lead to complete desiccation (runaway drying scenario; Waddington et al., 2015). Regardless of the actual mechanisms, the peatland demonstrates resilience in mitigating drying. Alongside warming-induced increases in plant productivity (Gallego-Sala et al., 2018), the input of recalcitrant and fire-resistant Sphagnum litter (Loisel and Yu, 2013), and increased acidity inhibiting microbial activity (van Breemen, 1995), surface wetting may contribute to maintaining or even enhancing the C sink and might be related to the high NECB observed in atmospheric C flux measurements (Fig. 5).

Figure 8 A conceptual diagram for the evolution of young, shallow, and ice-poor permafrost peatlands under climate change based on findings of this study.

5 Synthesis and conclusions

Through a detailed multi-core and multi-proxy anatomy of a typical ice-poor permafrost peatland in eastern Eurasia, we are able to understand how peatland formation, succession, and surface conditions responded to past climate variability and identify the roles of permafrost dynamics in regulating these responses. We summarize the main findings as a conceptual diagram describing the evolution of young, shallow, and ice-poor permafrost peatlands under climate change and into the future (Fig. 8). The general conclusions and broad implications are as follows:

1. Permafrost aggradation over permeable mineral-rich soils or substrates is a potentially important but overlooked agent in the formation of new peatland areas by maintaining poorly drained, waterlogged conditions. This permafrost–hydrology feedback highlights the need to consider surficial geology and cryosphere dynamics in understanding the genesis of global peatlands.

2. Ice-poor permafrost peatlands exhibit weak sensitivity to natural climate variability but clear regime-shift behavior under rapid and large-magnitude anthropogenic climate warming. Unlike ice-rich permafrost peatlands with highly dynamic surface morphology, ice-poor permafrost peatlands are characterized by a high degree of stability and homogeneity. The simple succession without involving thermokarst collapse means that their future greenhouse gas emissions and radiative forcing should be substantially smaller than those predicted by models developed for ice-rich permafrost peatlands (Hugelius et al., 2020).

3. Several internal feedback mechanisms may be involved in maintaining the hydrology and C accumulation of ice-poor permafrost peatlands under ongoing climate change. These behaviors reflect peatlands functioning as complex adaptive systems (Belyea and Baird, 2006).

4. Our analysis combining present-day atmospheric C flux measurements presents empirical evidence that similar ice-poor permafrost peatlands in eastern Eurasia are persistent C sinks. Simple protection measures such as fire prevention programs currently being implemented in northern Northeast China will be effective in preserving these significant terrestrial C stocks in the future.

Appendix A: Macrofossil diagram of peatland cores

Figure A1 Macrofossil diagram of core TQ22-C1.

Figure A2 Macrofossil diagram of core TQ22-C2.

Figure A3 Macrofossil diagram of core TQ22-C3.

Figure A4 Macrofossil diagram of core TQ22-C4.

Figure A5 Macrofossil diagram of core TQ22-C5.

Figure A6 Macrofossil diagram of core TQ22-C6.