Recent peat and carbon accumulation on changing permafrost landforms along the Mackenzie River valley, Northwest Territories, Canada

Northwestern Canada is currently warming nearly four times faster than the global average, driving accelerated permafrost thaw and changes to ecosystem vegetation, hydrology and landscape structure across the landscape. While permafrost peatlands constitute a large carbon reservoir, there is no consensus yet on the direction and magnitude of changes to their vulnerable carbon balance. Here, we assessed changes in peatland ecosystems following permafrost thaw at three sites located along a 1000 km long climate and permafrost gradient along the Mackenzie River valley, Canada. Specifically, we examined vegetation succession over the last few decades to evaluate the possible impact of climate warming on peat and carbon accumulation. Results from the palaeoecological analysis of 20 surficial peat cores, supported by robust chronologies, show a return to Sphagnum accumulation since ca. 1980 CE in the sporadic and discontinuous permafrost zones and ca. 2000 CE in the continuous permafrost zone. The average rates of peat and carbon accumulation reached 4 mm yr−1 and 134 g C m−2 yr−1 at the northernmost site in the continuous permafrost zone. In contrast, peat and carbon accumulation reached 3 mm yr−1 and 81 g C m−2 yr−1, respectively, in the sporadic and discontinuous permafrost zones. This study highlights the need for a net carbon budget that integrates the recent accelerated Sphagnum growth and carbon uptake from the atmosphere to better assess the potential carbon emissions offset following permafrost thaw. High-resolution palaeoecological studies can offer insights into decadal-scale patterns of vegetation and carbon balance changes to improve model predictions of peat climate-carbon cycle feedbacks.


Introduction
Northwestern Canada is currently warming nearly four times faster than the global average (Rantanen et al 2022). The mean annual air temperature in Canada has increased by 1.7 • C since 1950, and near-surface warming in northwestern Canada is reaching 3 • C in Arctic and subarctic latitudes (Bush and Lemmen 2019). The recent increase in air temperature has accelerated the degradation of ice-rich permafrost landforms (perennially frozen ground for >2 years), leading to the gradual or abrupt release of previously frozen carbon to the atmosphere as carbon dioxide (CO 2 ) and methane (CH 4 ) (Schuur et al 2015, Turetsky et al 2020. In Canada, peatlands store ∼150 Gt carbon (C), of which around 80% is stored in permafrost, hence representing one of the largest global C permafrost stocks (Sannel and Kuhry 2008, Tarnocai et al 2011, Treat et al 2016. The permafrost distribution in Canada varies from isolated (<10% areal extent), sporadic (⩾10%-50%), and discontinuous (>50%-90%) to continuous r(>90%-100%), corresponding to regions where the mean annual air temperature varies from −4 • C to −10 • C (Brown 1963, Gruber 2012. While the degradation of ice-rich permafrost has been observed since the beginning of the twentieth century (Quinton et al 2019), there is no consensus yet on the direction and magnitude of the net carbon sink-source strength of peatland ecosystems affected by degrading permafrost (Gallego-Sala et al 2018). Some studies have presented the fact that a warmer climate and associated permafrost thaw may increase the release of CO 2 and CH 4 to the atmosphere (Dorrepaal et al 2009, Schuur et al 2015, Helbig et al 2017a, 2017b, Turetsky et al 2020. Meanwhile, other studies have suggested that climate warming may increase the growing season length and primary productivity, both promoting vertical peat accumulation and lateral peat expansion on decadal to centennial scales (Charman et al 2013, Ruppel et al 2013, Magnan et al 2022. Continued climate warming and the associated thaw of ice-rich permafrost near the southern limit of permafrost have already transformed some forested landscapes into thermokarst lakes and wetland ecosystems in the Taiga Plains (Quinton et al 2009, Quinton et al 2019, Wright et al 2022. There, thaw-induced land cover changes have led to the increasing hydrological connectivity of the fragmented landscape comprising saturated, partly inundated wetlands Quinton and Baltzer (2013). The associated shift from aerobic to anaerobic decomposition leads to more CH 4 emissions into the atmosphere (Helbig et al 2017b). Over time, the disconnected landscape drains and transforms into interconnected wetlands (Connon et al 2015, Haynes et al 2018. As the air temperature rises, permafrost thaw occurs both vertically (i.e. thickening of the active layer) and laterally, ultimately leading to the development of permafrost-free thermokarst wetlands and the degradation of the surrounding permafrost peat plateaus (McClymont et al 2013, Pelletier et al 2017, Carpino et al 2021. Thaw-induced degradation occurring along the edges of the forested peat plateaus leads to subsidence with inundated margins and a return to Sphagnum accumulation (Baltzer et al 2014, Pelletier et al 2017. In the continuous permafrost zone, permafrost thaw occurs mostly vertically in peatlands with a deepening of the active layer; however, its degradation has not been fully documented (Rey et al 2020).
The number of studies documenting recent peat accumulation in response to permafrost thaw remains quite low (Lamarre et al 2012, Loisel and Yu 2013, Heffernan et al 2020, Larson et al 2022, Sim et al 2022. Following degrading permafrost, the detailed succession from ligneous peat to Sphagnum peat was first documented by Lamarre et al (2012) in Nunavik, Québec. A recent study of a large high-resolution dataset covering the last century and spanning 33 sites across northern Québec has synthesized the recent vegetation state-shift over recent decades (Magnan et al 2022). As such, the shift to Sphagnum-dominated peat was found to be more resistant to hydrological changes and represents an important control on carbon dynamics following recent thaw-induced peat accumulation (Estop-Aragonés et al 2018, Larson et al 2022, Sim et al 2022.
As thermokarst landforms develop, the dominant vegetation on permafrost peat plateaus and their thawing edges, i.e. black spruce (Picea mariana) and different shrub species, is subject to collapse and followed by a shift to Sphagnum mosses with high biomass productivity (Heffernan et al 2020, Larson et al 2022. In eastern Canada, the shift from ligneous to Sphagnum-dominated peatland ecosystems, characterized by more decay-resistant mosses, such as Sphagnum fuscum and Sphagnum capillifolium, suggests an increase in carbon accumulation since ca. 1950 (Magnan et al 2022). Some climate models have recently included permafrost carbon feedback in their projections (Miner et al 2022), but without considering a potential increase in peat accumulation on permafrost landforms. Evaluating the response of highlatitude peatland ecosystems to a rapidly changing climate can therefore have major global implications considering their fundamental role in the global carbon cycle (Ciais et al 2014, Winkler et al 2019, Miner et al 2022. Using a palaeoecological approach, the goal of this study was to quantify the recent peat and carbon accumulation following permafrost thaw along a 1000 km climate and permafrost gradient in the Mackenzie River valley in the Taiga Plains ecozone of the Northwest Territories, Canada. To reach this goal, our objectives were to: (1) characterize patterns in recent peat accumulation over permafrost landforms in three peatlands of the sporadic, discontinuous and continuous permafrost zones; (2) reconstruct the vegetation succession following permafrost thaw; and (3) quantify changes in recent peat and carbon accumulation rates on permafrost landforms since 1950.

Study sites and sampling
Three study sites in the Mackenzie River valley of the Taiga Plains ecozone were selected, from south to north along a climate and permafrost gradient (figures 1 and 2): Scotty Creek (SCC), Smith Creek (SMC) and Havikpak Creek (HPC). The regional climate is continental with long, cold winters and short, cool summers (Wright et al 2022). Following the deglaciation of the Laurentide Ice Sheet around 11 ka BP (Lemmen et al 1994), cooling conditions led to permafrost aggradation around 6-5 ka BP in the northern part of the Taiga Plains known as the Peel Plateau region (Zoltai 1995, Vardy et al 1998), and around 4-3 ka BP in the Great Slave Lowlands Tarnocai 1975, Zoltai 1993). Poorly drained lacustrine deposits from the postglacial Lake McConnell, as well as a low relief, led to the establishment of thick organic deposits and development of extensive peat landscapes (Wright et al 2022).
Prior to the sampling field campaign, peatlands were characterized using imagery available on Google Earth. The study sites were selected based on the permafrost extent (i.e. sporadic, discontinuous and continuous), trophic regime (i.e. ombrotrophic) and peatland size (figure 2). Five peatlands were preselected in each permafrost zone and three were visited in the field based on their regional representativeness, size and access to the road (<1 km) before the final site selection was made. In each peatland, 5-10 m long lateral transects were defined on the permafrost peat plateaus from the thawing edge of the adjacent thermokarst wetland to the end of each transect. Cores were retrieved at systematic 1 or 2 m intervals along the transects. A total of 20 peat cores (SCC: n = 7, SMC: n = 6, HPC: n = 7) were collected in September 2019 for SCC and SMC, and June 2019 for HPC, with depths varying between 19 and 38 cm (see the supplementary material text S.1 for a detailed description of the coring methods and study sites).

Climate data and analysis
Time series from 1950 to 2017 for the three study sites were extracted from the homogenized climate data product provided by the ESCER Center (Centre pour l'étude et la simulation du climatà l'échelle régionale) of the Université du Québecà Montréal (UQAM). Trends in climate data were derived from gridded homogenized climate data products  from Vincent et al (2018) 1950-2017, 1950-1980 and 1980-2017, a Mann-Kendall test was performed on each time series with a Sen's slope of 10 for a decadal trend (Mann 1945, Kendall 1975). Finally, a changepoint analysis was conducted on MAAT ( • C), mean total annual precipitation (MTAP, mm) and GDD 0 (>0 • C, growing degree days) for the period of 1950-2017 to identify significant changes in climate trends (α = 0.05).

Laboratory analyses 2.3.1. Bulk density, organic/mineral content and C:N ratios
Organic matter density (g cm −3 ), carbon density (g C cm −3 ) and mineral content (%) were estimated using the modified loss-on-ignition protocol from Chambers et al (2011) at a systematic resolution of 1 cm. Carbon density was then determined by multiplying each value of organic matter density by 0.5 (sensu Turunen et al 2002). C:N ratios were measured on each peat core at 2 cm intervals using a Carlo Erba NC 2500 elemental analyzer at the Stable Isotope Geochemistry Laboratory, GEOTOP-UQAM.

Plant macrofossil analysis
Plant macrofossil analyses were performed at 4 cm intervals from 3 cm 3 subsamples, as in Mauquoy et al (2010). When a change in bulk density was identified, horizons were analyzed at narrower intervals of 1 or 2 cm. To thoroughly assess the dynamics of recent shifts in vegetation assemblages, a transition horizon in which Sphagnum was introduced was identified as part of the macrofossil analyses (figure 3). Macrofossil assemblages preceding this transition horizon were characterized by peat dominated by >80% ligneous fragments with very low percentages of Sphagnum (<1%-5%). This ligneous horizon was identified in 15 of the 20 cores, topped by a Sphagnum assemblage, mainly represented by S. fuscum, Sphagnum rubellum and S. capillifolium. The ecosystem state-shift was confirmed when Sphagnum spp. exceeded 50% of the macrofossil assemblage.

Dating and chronologies
Chronologies were determined by pairing radiocarbon ( 14 C) and radiogenic lead-210 ( 210 Pb) dating. A total of 52 organic samples were submitted for 14 C to the A.E. Lalonde AMS radiocarbon laboratory at the University of Ottawa and processed into a mass spectrometry accelerator. The main stratigraphic horizons within the 20 peat cores were targeted for radiocarbon dating and, more specifically, before and after the Sphagnum shift. Sphagnum fragments were prioritized for dating. Radiocarbon samples were calibrated using the IntCal20 calibration curve (Reimer et al 2020), and modern dates were calibrated using the NHZ1 post-bomb atmospheric radiocarbon curve (Hua et al 2013). Lead-210 analysis was performed on 1-2 cm intervals for the most recent horizons (<150 years) at the GEOTOP-UQAM Radiochronology Laboratory using alpha spectrometry. Combined 14 C and 210 Pb data allowed us to produce age-depth models on all 20 peat cores using the Bayesian Plum model, rplum v0.1.4 (Aquino-López et al 2018) in the R v3.6.3 (R Core Team 2020). For cores that did not reach the supported 210 Pb, prior distribution was applied from other cores at the same sites (supplementary material, table S.5).

Accumulation rates
Apparent carbon accumulation rates (ACAR, g C m −2 yr −1 ) were calculated from dry bulk density (g m −3 ) for all 20 cores using two time periods: pre-1980 and post 1980. The year 1980 was chosen as it represents the average year for climatic changes identified from the changepoint analyses. ACARs were then calculated for 1950-2017(ACAR 1950-2017), 1950-1980(ACAR 1950-1980) and 1980-2017(ACAR 1980-2017 to allow for comparisons within each time period at each of the three sites. The dates used were those obtained by the age-depth models. Peat accumulation rates (PARs, mm yr −1 ) were also calculated for the same time and transition periods. ACARs (gC m −2 yr −1 ) for 1950-2017, 1950-1980 and 1980-2017 were analyzed using a oneway ANOVA and a Tukey post-hoc test to evaluate the accumulation differences in-between the three sites.  4). The ANOVA test highlighted p-values smaller than 0.05 for all three sites; therefore, SCC, SMC and HPC are significantly different from each other. A significant change in MAAT was identified around 1985 at SCC, and around 1993 at SMC and HPC (supplementary material, figure S.6). The results for the MTAP showed a decrease from south to north (SCC: 363 mm, SMC: 306 mm, HPC: 252 mm) but also within each site, suggesting that the Taiga Plains is subjected to a general drying tendency, although statistical tests showed no significant trends. SCC and HPC were both marked by recent changepoints (2013 and 2006, respectively). With regard to the growing season, the GDD 0 results were not synchronous between the three sites but showed a recent increase within each site, and thus a lengthening of the growing seasons. The changepoint was much more recent in SCC (2012) and the trend was more pronounced compared to SMC (1987) and HPC (1987) (supplementary material, figure S.6).

Chronologies and vegetation succession 3.2.1. Scotty Creek
Of the seven cores at SCC, two were entirely composed of Sphagnum peat (SCC_LNY-X), without basal ligneous peat. For the other five cores (SCC_LNZ-V, SCC_LSZ-X-V), the age of the basal ligneous section varied between 1785 ± 60 CE and 1872 ± 59 CE, while the Sphagnum peat transition was dated around 1836 ± 33 CE and 1937 ± 11 CE. The PARs and ACAR values of the ligneous horizons were low (respectively 0.7 mm yr −1 and 37.8 gC m −2 yr −1 ). Decomposition was high, as suggested by low C:N ratios ranging from 40 to 70. In most cores, Polytrichum sp. mosses were found in the top layers of the ligneous peat.
The transition period was identified by the succession from ligneous peat to Sphagnum-dominated peat ( figure 3). This transition is characterized by the introduction of S. fuscum and S. rubellum with some herbaceous species within the macrofossil assemblages. Supported by the detailed chronologies (supplementary material, table S.2), the transition period identified in the northern transect of SCC may have lasted between 30 and 60 years, while the transition period was much shorter in the southern transect (3-15 years) before it shifted to a Sphagnum-dominated ecosystem. The shift to a Sphagnum-dominated environment varied between 1898 ± 42 CE and 1967 ± 5 CE in the northern transect, and between 1992 ± 7 CE and 1998 ± 2 CE in the southern one, thus more recent and more rapid.
Following the vegetation shift, high percentages (>80%) of S. fuscum and Sphagnum angustifolium were identified in the peat profiles. The PARs and ACARs fluctuated around 3.5 mm yr −1 and 84 gC m −2 yr −1 , respectively, corresponding to the most recent and poorly decomposed accumulation. Poor decomposition was also confirmed by high C:N ratios ranging from 90 to 160. The recent Sphagnum shift in the northern transect occurred between ca. 1898 ± 42 CE and 1967 ± 5 CE, the earliest change being recorded near the thawing edge of the thermokarst wetland (figure 5). Conversely, along the southern transect, the Sphagnum shift occurred almost synchronously during the last few decades (ca. 1995 ± 5 CE near the collapse scar, and ca. 1992 ± 7 CE further in the forested permafrost peat plateau). PARs and ACARs were also calculated according to each transition unit individually (supplementary material, table S.1).

Smith Creek
Three of the six cores analyzed at SMC were entirely composed of Sphagnum peat (SMC_LNZ-X and SMC_LSX) while, in SMC_LNV, SMC_LSV and SMC_LSZ, the ligneous horizon was reached and thus identified at the bottom of the core. The base of the ligneous horizon was dated between 1789 ± 56 CE and 1910 ± 15 CE, while the shift to a dominance of Sphagnum peat was dated between 1821 ± 59 CE and 1979 ± 8 CE. The PARs and ACARs of the welldecomposed ligneous horizon were low (0.3 mm yr −1 and 41.5 gC m −2 yr −1 , respectively) and C:N ratios ranging from 30 to 80 suggested the high degree of decomposition. Apart from unidentified ligneous fragments, some fragments of Polytrichum sp., Pohlia sp. and Cladonia sp. were found in the macrofossil assemblages of the ligneous horizon.
The transition period from the ligneous horizon to a dominance of Sphagnum mosses was characterized by the presence of S. fuscum, S. capillifolium and Sphagnum warnstorfii. As two of the three cores of the northern transect were entirely composed of Sphagnum peat (SMC_LNX and SMC_LNZ), it was not possible to determine any transition along the northern transect (figure 6). Only SMC_LNV suggested a shift around 1835 ± 51 CE. Along the southern transect, the shift to a Sphagnum-dominated environment in cores SMC_LSV and SMC_LSZ was identified between 1985 ± 5 CE and 1991 ± 5 CE, with a short transition time of approximately 12 and 19 years.
The vegetation succession following the Sphagnum takeover was characterized by a dominance of S. fuscum and S. rubellum. The PARs and ACARs varied between 3.2 mm yr −1 and 66.7 gC m −2 yr −1 respectively, which corresponded to the poorly decomposed peat, as confirmed by high C:N ratios ranging from 60 to 150. The Sphagnum takeover occurred approximately 185 years ago near the collapse scar (SMC_LNV). Along the southern transect, the Sphagnum takeover occurred around 1985 ± 5 CE near the thawing edge of the related expanding thermokarst wetland. PARs and ACARs were also calculated for each transition unit individually (supplementary material, table S.2).

Havikpak Creek
In contrast to SCC and SMC, peat stratigraphy of the HPC cores showed more heterogeneity in the macrofossil assemblages ( figure 7). The base of the ligneous horizon was dated between 1875 ± 45 CE    and 1964 ± 12 CE, and the ligneous horizon was found to be often mixed with some brown mosses, such as Aulacomnium palustre, Aulacomnium turgidum and Polytrichum sp. The PARs and ACARs of the ligneous horizons ranged between 2.3 mm yr −1 and 105.8 gC m −2 yr −1 , respectively, with C:N ratios varying from 50 to 110. The shift to Sphagnumdominated peat was identified between 1936 ± 23 CE and 2001 ± 4 CE.
The transition period was characterized by the introduction of S. fuscum, S. angustifolium and Sphagnum balticum together with some herbaceous and brown mosses. This transition period was relatively short before it shifted to a Sphagnumdominated ecosystem, i.e. between 12 and 26 years (i.e. between 1994 ± 5 CE and 2008 ±10 CE) in the northern transect and 4-22 years (i.e. between 1958 ± 9 CE and 2006 ± 6 CE) in the eastern transect.
Following the transition period, the vegetation assemblages in the cores were dominated by high percentages of S. fuscum and S. balticum together with Anastrophyllum minutum, Barbilophozia sp. and Leiomylia anomala. The PARs and ACARs in this Sphagnum-dominated section varied around 5.2 mm yr −1 and 138 gC m −2 yr −1 . The poorly decomposed conditions were confirmed by high C:N ratios ranging from 70 to 170. The transition period lasted approximately 11-25 years on the northern transect and 13-61 years on the eastern one. The 14 C and 210 Pb dates suggest that the shift to a Sphagnumdominated ecosystem was first initiated at the edge of the bog and expanded laterally into the forested margins. PARs and ACARs were also calculated according to each transition unit individually (supplementary material, table S.2). Across all three sites, the values for PAR 1980-2017and ACAR 1980-2017 almost doubled compared to those for PAR 1950-1980and ACAR 1950-1980 , roughly corresponding to the recent period ( figure 4). However, both PAR 1980-2017and ACAR 1980-2017 were generally lower at SCC and SMC than at HPC. Site-average ACAR 1950-2017, ACAR 1950-1980and ACAR 1980-2017 were calculated for each site. The oneway ANOVA and Shapiro-Wilk test showed that SCC and SMC were not significantly different from each other (p-value > 0.05) over all three climate periods. Thus, both SCC and SMC showed similar PAR and ACAR trends, both different for HPC. The ANOVA showed that: (1) PAR 1950-2017and ACAR 1950-2017 are increasing from south to north, with HPC showing higher PAR 1950-2017and ACAR 1950-2017 ; and (2) the peat and ACARs were also higher in the northernmost latitude for the 1950-1980 and 1980-2017 periods (figure 9).

Recent climate-change-induced permafrost thaw
Our results suggest that a significant lengthening of the growing season occurred earlier in SCC and SMC (ca. 1987) than in HPC (ca. 2003). In our homogenized climate data, no increase in total annual precipitation was detected across the climate and permafrost gradient over the periods 1950-2017 and 1980-2017. At the same time, a widespread and rapid thaw of relatively warm and ice-rich permafrost has been observed in the Taiga Plains (Burn et al 2009, Wright et al 2022. At the sporadic permafrost (SCC) and discontinuous permafrost sites (SMC), permafrost degradation was initiated nearly 50-60 years ago (Wright et al 2022), which coincides with our reconstructed changes in vegetation succession and peat accumulation. Both sites are characterized by expanding permafrost-free thermokarst wetlands surrounded by thawing permafrost peat plateaus (Connon et al 2018,  1950-2017, 1950-1980 and 1980-2017: n is the sample size, and a is significantly different from b (p-value < 0.05).
Quinton et al 2019). Peat plateau edges in the sporadic and discontinuous permafrost zone are more sensitive to ground subsidence due to vertical loss and lateral permafrost degradation induced by heat transfer (Baltzer et al 2014, Haynes et al 2018. With permafrost degradation, peat plateaus subsided, forming new Sphagnum-dominated wetlands (Myers-Smith et al 2008, Quinton et al 2011, Haynes et al 2018. On the permafrost peat plateaus of SCC and SMC, the detailed chronologies obtained from the different cores also suggest a return to patchy Sphagnum accumulation over previously frozen landforms ten years before the peatland at HPC.

Shift in permafrost landform vegetation composition in response to climate-change-induced permafrost thaw
The results of the macrofossil analyses from Sphagnum cores allowed us to determine a systematic succession of three main vegetation assemblages associated with the permafrost dynamics of the Taiga Plains: a dominance of ligneous vegetation when permafrost was more stable, a transition period with the introduction of some Sphagnum species with the beginning of permafrost thaw and, finally, a shift to a dominance of Sphagnum species (mainly S. fuscum and S. capillifolium) over a low or absent frost table. These three main vegetation assemblages were observed throughout all three sites along the Mackenzie River valley, but the initiation of the vegetation shifts is not due to the same permafrost thaw dynamics nor lasted the same number of years (figure 10). As the results of peat and carbon accumulation showed, the sporadic and discontinuous permafrost coring sites (SCC and SMC) suggest relatively similar permafrost thaw mechanisms and vegetation successions, setting the continuous permafrost site of HPC apart.
On the permafrost peat plateaus, the ligneous vegetation, corresponding to a 'low-accumulation period' , as described in Sannel and Khury (2008), Lamarre et al (2012) and Pelletier et al (2017), was followed by a transition period, corresponding to a time where thaw initiated at SCC and SMC. Interpreted as the shift from a ligneous horizon to Sphagnum-dominated peat, the transition period is also consistent with a change in climate conditions, such as warming temperatures (0.48 • C to 1.02 • C decadal increase in MAAT since 1980, figure 4) and lengthening of the growing season since 1990 (see section 3.1). At SCC and SMC, the transition period from approximately 1835 ± 33 CE and 1828 ± 38 CE to 1990 ± 6 CE was captured by a succession from a forested environment to the development of several peat hummocks in the degraded peatland margins. This shift was supported in the macrofossil assemblages by the presence of S. fuscum, Polytrichum sp., Rhododendron groenlandicum and some Picea needles, reported as typical of degrading peat plateaus (Sannel and Kuhry 2008). The post-thaw period at SCC and SMC, characterized by a shift to dominating ombrotrophic Sphagnum species, suggests an increase in soil moisture associated with permafrost thaw. The interpretation of this period was supported by a return to Sphagnum accumulation with high values of PAR and ACAR, coinciding with permafrost thaw and degrading peat plateaus. Furthermore, freeze and thaw periods were also identified in the topmost horizons of the cores and confirmed by the alternations of S. fuscum and Polytrichum sp. in the macrofossil assemblages (Sannel and Kuhry 2008). Spatio-temporal differences in vegetation succession patterns were also found within each site. At SCC and SMC, the initiation of the transition periods was found to be more or less synchronous in-between the southern and northern transects of each site, while HPC showed a clear gradual return to Sphagnum accumulation on the eastern transect, as presented in figure 10.
In all three sites, the pre-thaw period was characterized by ligneous vegetation assemblages similar to those described at other sites in northwestern Canada within the continuous permafrost zone (Zoltai 1993, Robinson and Moore 2000, Oksanen 2006, Loisel et al 2014. This pre-thaw period lasted until ca. 2001 CE or until the beginning of permafrost thaw. With regard to the transition period, HPC showed a much more recent response to permafrost thaw and increasing air temperatures (from 1935 ±23 CE to 2010 ± 9 CE on average) than SCC and SMC. Accumulation rates were higher during the transition period of HPC than SCC and SMC, with much younger shifts in vegetation succession (ca. 1980 ± 7 CE on average). The post-thaw period or shift to a dominance of Sphagnum in HPC was marked by high PARs and ACARs. What sets the continuous permafrost site (HPC) apart from the other sites is the gradual temporal shift in Sphagnum accumulation from the peatland to the margins, as expressed in figure 10. These results suggest that HPC is undergoing both lateral peat expansion (transect east) and vertical peat accumulation (transect north). In this study, the continuous permafrost site of HPC shows the highest accumulation rates for the last 37 years of the entire Mackenzie River valley transect of the Taiga Plains that were reconstructed in this study.
A previous study realized more than 20 years ago in the continuous permafrost zone close to HPC, suggested that peat and ACARs were found to decrease with lower MAAT (Robinson et al 2003). However, our results showed that with the recent rapid warming increase, peat and carbon accumulation have increased with latitude. We believe that, since the sporadic-discontinuous permafrost area has experienced critical thaw for a longer period of time (over 50-60 years, Wright et al 2022), Sphagnum peat may also have undergone some decomposition processes affecting the net values of accumulation rates (Myers-Smith et al 2008, Gallego-Sala et al 2018. However, the high increase in MAAT at HPC probably influenced rapid permafrost thaw with longer growing season lengths, hence positively influencing vegetation productivity. These important recent and rapid changes suggest that the northernmost latitudes of the Taiga Plains are experiencing the most notable ecological changes since 1980, and highlight the role of heterogeneous landscapes and site-specific thaw patterns in scaling-up studies.

Conclusion
The permafrost peatlands of northwestern Canada are experiencing important changes as this region is warming up to four times faster than the rest of the planet. Previous studies found that the permafrost regions of northwestern Canada have been degrading rapidly since 1950. The results of the present study suggest that with increasing air temperature, lengthening of the growing seasons and recent permafrost thaw, peatlands of the Mackenzie River valley are showing a recent return to Sphagnum accumulation and some potential lateral expansion influenced by local topography and changing hydrological states, allowing the Sphagnum-dominated vegetation shift to increase carbon uptake. Using a palaeoecological approach, analyses of 20 surficial peat cores spread across three study sites and supported by robust chronologies showed that: (1) A shift to patchy Sphagnum-dominated peat occurred on thawing permafrost peat plateaus of the sporadic-discontinuous permafrost zones and thawing peatland margins of the continuous permafrost zone, allowing vertical and lateral Sphagnum peat expansion and an increase in carbon accumulation; (2) With the return of Sphagnum over the lichenous permafrost peat plateaus, rates of peat and carbon accumulation are higher in the northernmost site, reaching up to 4 mm yr −1 and 134 g C m −2 yr −1 , respectively; (3) Increased patchy Sphagnum peat and ACARs occurred around 1980 CE following permafrost thaw, an increase in the MAAT and lengthening of the growing season.
This study shows how a palaeoecological approach can improve our understanding of changing permafrost landforms across northern latitudes. It also highlights the need for an integrative carbon budget that includes not only annual CO 2 and CH 4 release from permafrost thaw but longer-term carbon uptake from the recent Sphagnum growth that has not yet been clearly captured by in situ measurements. This study was defined to specifically quantify the recent return of Sphagnum accumulation on forested permafrost landforms. The important role of increased photosynthesis should be considered in future model projections to better assess the response of northern latitude peatlands to climate change.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).