We collected cores for the mesocosm incubations from a palsa mire located in a discontinuous permafrost area in the Finnish part of Sápmi in March 2023. The Peera palsa mire (68.8778° N; 21.0792° E) is composed of four palsas , surrounded by fens. The mean annual air temperature measured at the Kilpisjärvi station is -1.7 °C, and the average annual precipitation (rain and snow) is 546 mm (1991–2020, FMI, 2024). In this study, we focused on the southernmost landform (Fig. ). Over more than 60 years of aerial imagery monitoring, the palsa has lost 77 % of its surface area, transitioning into thermokarst bog and fen habitats. This transition is associated with a shift in vegetation from shrub-dominated peat to open water, followed by colonization by Sphagnum mosses and sedges . Based on the definitions from , the palsa is surrounded by three developing thermokarst ponds and four emerging ponds at its edge. The thermokarst ponds are located at the same elevation as the surrounding fen, which lies approximately two meters lower than the palsa. We measured a peat thickness of 1.7 m and an active layer depth of 60 cm at the top of the Peera palsa dome. The active layer was measured during the first exploratory expedition in October 2022. Since this was at the end of the growing season, we assume this represents the maximum annual active layer depth. The vegetation at the sampling locations on Peera palsa is characterized by shallow-rooted species such as Betula nana subsp., Empetrum nigrum subsp., and Vaccinium myrtillus, while the surrounding fen is dominated by deep-rooted vegetation such as Eriophorum, Sphagnum, and sedges (Fig. S1 in the Supplement).

We sampled cores from two thaw stages: the intact palsa and the thermokarst fen (Fig. ). To minimize soil disturbance, sampling was conducted in winter (March 2023), when frozen conditions help preserve water content and prevent artificial compaction. We collected twelve cores (1 m long, 7.5 cm diameter) from the Peera palsa and three from the fen (Fig. ). At both sites, we also collected a “long core” down to mineral soil for sedimentary analysis. Mesocosms from the palsa were drilled using a snow, ice, and permafrost (SIPRE) auger (Jon's Machine Shop, Alaska, USA). In the fen, only the top 15 cm were frozen, so we used the SIPRE auger for that layer and then drove a metal pipe into the unfrozen soil to complete the core. To preserve soil structure, the pipes were frozen vertically. We wrapped the palsa samples and the frozen top layers from the fen in core foil and stored them in thermoboxes.

We collected 45 L of water from a thermokarst pond adjacent to the palsa (68.8775° N; 21.0798° E), using pre-cleaned 10 L canisters. We drilled through the surface ice with the SIPRE auger before sampling.

Samples were stored at -20 °C at the Kilpisjärvi Research Station during the expedition. They were later shipped frozen to AWI in Potsdam, Germany, using a refrigerated truck and stored at -20 °C until the start of the experiment.

We characterized the thawing response by determining shifts in the emissions over time. To do so, we applied a changing-point-based method coupled to a sliding mean . When a potential shift is detected, a Student's t-test is run to validate whether the anomaly is statically different from the mean value. We tested the method for several cut-off lengths to cover short-term and long-term regimes. In the end we chose a cut-off length of 5 d. We ran the Rodionov test with the rshift package .

We tested the differences among treatments and thaw stages for the averaged fluxes with a one-way ANOVA coupled to a TukeyHSD test. We also grouped the treatments between flooded and dry conditions and tested for significance. Before performing statistical test, we tested the data for independency, normality and homogeneity of variance. Finally, to ensure that filtering did not affect the statistics, we performed a Kruskal-Wallis test followed by pairwise comparisons using Dunn's test on both the non-filtered (before quality control) and filtered (after quality control) datasets. Because extreme values can strongly influence mean-based comparisons, we employed robust, median-based statistical methods that are less sensitive to outliers to compare the two datasets (Table S4).

We used R for all the computations.

pH values in the active layer ranged from 3.4 to 3.6. The pH in the permafrost layer was less acidic, with values between 4.2 (depth 100 cm) and 4.9 (permafrost table). EC was three times higher in the active layer than in the permafrost layer. As with pH and EC, the DOC concentration increased in the permafrost layer (ranging from 120 to 165 mg L⁻¹ in the active layer and from 402 to 480 mg L⁻¹ in the permafrost layer) (Fig. ). The pH and EC in the fen core were homogeneous along the soil profile, with pH values similar to those in the permafrost layer and EC values of the same order of magnitude as those in the active layer. Unlike in the palsa, DOC concentration in the fen decreased with depth (Fig. ). TOC and C : N ratios did not differ between the two sites. TOC values ranged from 41 % to 52 %, with the lowest values (41 %) measured at the permafrost table. The C : N ratios remained consistent along the soil profile (between 30 and 40), with a decrease at the permafrost table and at a depth of 70 cm (16) (Fig. ).

Overall, we observed a difference between the active and permafrost layers in the soil pore water, while the sediment parameters remained stable along the soil profile. The soil parameters in the fen were mostly stable with depth. Finally, the difference between the palsa and the fen was mainly evident in the composition of the soil pore water.

The FTIR spectra of the palsa samples collected at 0–2 and 18–22 cm depth showed band intensities 15 % (0–2 cm) to 30 % (18–22 cm) higher for the CH and C=O bands (Fig. a – 0 and 20 cm), compared to the fen samples. At 60 cm depth (the permafrost table), the opposite trend was observed (Fig. a – 60). No differences were observed in the FTIR spectra of palsa and fen samples collected at 100 cm (Fig. S12). After incubation, the strongest decomposition of OM was observed in samples from the permafrost table (60 cm), with a DI of 1.8 (Fig. b). This was also supported by an increase in the band intensities of CH and C=O groups post-incubation (Fig. b). The DI ratios in the flooded (abrupt thaw simulation) cores were approximately 1.7 times higher than those in the Control (Figs. b; S12). Little changes were observed in other layers. In the fen cores, the most significant increase in band intensity post-incubation was detected in the bottom layer (100 cm), which also had the highest DI value of 1.9. Above this depth (0–80 cm), little to no change was observed during incubation.

Under in-situ conditions (i.e., prior to incubation), we measured a higher degree of OM decomposition in the active layer of the palsa compared to the fen. After incubation, the depth of greatest OM decomposition varied by site: in the palsa, the permafrost table experienced the highest decomposition, whereas in the fen, most changes in CH and C=O bands occurred in the bottom layer.

The palsa mesocosms exhibited a net uptake of CH₄ for all treatments (ranging from -3.2±0.6 to -34.6±15.4; respectively Control 0–80 and Non-Filtered 0–100). Neither the thaw stages nor the treatments had a discernible effect on the CH₄ uptake rates (TukeyHSD, p-value > 0.05; Table ).

For the fen, we observed a burst of CH₄ emissions during the first week of incubation, followed by emissions close to zero (Fig. S15). This resulted in an average CH₄ emission of 9.7±17.8 µg-CH₄ m⁻² h⁻¹ during the first thaw stage (0–60). The second thaw stage (0–80) triggered another burst of CH₄ emissions for three days, while no change in CH₄ emissions was observed after the third stage (Fig. S15). In the fen, the average CH₄ emissions for each thaw stage indicated a continuous increase in CH₄ uptake throughout the incubation period (Table , Fig. S15).

During the incubation, we measured CO₂ emissions that were three to four times higher in the fen compared to the palsa when the cores were incubated under in-situ moisture conditions (Fig. a, Table ). Our results align with previous incubation studies showing two to four times higher CO₂ production rates from fens than from palsas . However, since we removed vegetation from the top layer, the CO₂ emissions we measured were likely higher than under true in-situ conditions, especially for the fens, which typically act as a C sink during the growing season . For comparison, found that seasonal CO₂ emissions were about three times higher under bare conditions than under vegetated conditions in a warmed treatment. Under control conditions, vegetated sites acted as a CO₂ sink during July and August, while bare sites remained a CO₂ source throughout the snow-free season.

The accumulation and decomposition of peat at the studied sites are shaped by distinct hydrological conditions. The fen site has been permafrost-free for approximately 25 years, in the uppermost 2.5 m (since 2000), during which the top 23 cm of peat has accumulated (Fig. ; ). This layer consists of modern C inputs from fresh vegetation. Under water-saturated conditions, decomposition is limited by the lack of terminal electron acceptors (TEAs) , which leads to a higher accumulation of OM . In contrast, the active layer of the palsa contained OM deposited over the last 4600 years and was exposed to decomposition due to dry conditions . This prolonged exposure to decomposition in the active layer of the palsa reduced the potential C loss from this site over time . Our FTIR results indicate that the OM in the active layer of the palsa, prior to incubation, is more decomposed than the OM in the first 60 cm of the fen, reflecting the different hydrological conditions (Fig. a) and OM inputs.

The dry soil conditions in the palsa led to a shrub-dominated environment with a dried Sphagnum layer at the surface. This type of vegetation is typically associated with acidic soils, (pH < 4), and our pH results (from 3.4 to 3.6) for the active layer fall into the range of magnitude of palsas . However, Sphagnum moss is difficult to decompose because it captures carbohydrates and releases organic acids that inhibit decomposition, which decreases the lability of OM . The shift from Sphagnum to sedge-dominated peat in the fen caused an increase in pH that facilitates OM degradation. Additionally, we measured DOC concentrations that were 2.5 times higher (up to 500 mg L⁻¹) in the first 20 cm of the fen core (Figs. , S21). DOC from sedge-dominated peat is more labile than DOC from shrub-dominated peat . Therefore, we observed a surface layer in the fen with higher and more labile DOC, which could explain the higher CO₂ emissions measured during the incubation.

To summarize, although aerobic conditions facilitate OM decomposition in the palsa, the high degree of peat decomposition combined with low OM inputs has resulted in more recalcitrant OM in the active layer of the palsa (Fig. a), potentially limiting further C mineralization . In contrast, while the fen is O₂-limited, the vegetation, fresh OM input, and adapted microbial communities led to threefold higher CO₂ emissions.

In our experimental setting, abrupt thaw simulations resulted in a 55 % increase in CO₂ emissions under flooded conditions compared to gradual thaw simulations (Control) (Table , Fig. ). These results indicate that the abrupt flooding scenario had a greater impact on CO₂ emissions than permafrost thaw alone. This contrasts with previous incubation studies simulating permafrost thaw in palsa, where lower CO₂ production was observed under anaerobic conditions . However, , in their mesocosm study, did not observe a significant difference in CO₂ emissions between flooded and in-situ moisture conditions. Under in-situ moisture conditions, the first 60 cm of the palsa remained aerobic. Therefore, even under flooded conditions, the meso-scale of our experiment delayed the establishment of anaerobic conditions compared to traditional incubations. After a three-month incubation, the flooded palsa mesocosms at 10 °C had not reached anaerobic conditions, which also explains the absence of CH₄ emissions. Under field conditions, chronosequence studies have shown that CH₄ emissions increase during the young and intermediate thaw stages (approximately 10–60 years after thaw), when soils are wetter and vegetation is adapted to these conditions. The lack of anaerobic conditions in our mesocosms likely reflects the short experimental duration compared to natural thaw dynamics (three months vs. several decades). As bogs mature and peat accumulates, the water table lowers, leading to drier conditions and a subsequent decline in CH₄ emissions . Additionally, higher CO₂ emissions under flooded conditions have also been observed in field studies during the first years after thaw (<4 years) , supporting the potential for increased CO₂ emissions with higher moisture content when O₂ remains available.

In their review paper, compiled studies from various experimental setups and soil types to explain the observed increase in CO₂ emissions under flooded conditions. They attributed this increase to higher C lability during flooding events, driven by several mechanisms: (1) changes in soil redox properties , (2) vertical substrate mobilization , (3) increased microbial biomass C , and (4) physical disruption, which makes soil OM more accessible . However, the latter mechanism is likely a minor contributor in peat soils, due to their low mineral content, which limits aggregate formation .

The sudden increase in soil moisture with flooding changes the availability of TEAs, changing the redox status and reducing the likelihood of CH₄ production. Some palsas are iron-rich environments, which can help protect OM from decomposition under aerobic conditions but under anaerobic conditions, Fe(III) can be reduced to aqueous Fe(II), thereby making OM more accessible for microbial decomposition . However, other more favorable TEAs such as NO3- could be used first to decompose OM and therefore delay the utilization of Fe(III).

Furthermore, flooding stimulates vertical transport within the soil column, mobilizing OM and nutrients from deeper layers to the surface, which can enhance C mineralization . In both scenarios, we observed the greatest increase in OM decomposition at the permafrost table (Fig. b). However, decomposition was more pronounced under flooding conditions (DI: 1.9 vs. 1.3 in flooded and control treatments, respectively; Fig. b). With enhanced vertical transport, freshly decomposed substrates can be mobilized to upper soil layers .

Rewetting soil can also lead to microbial cell lysis, releasing nutrients and increasing microbial biomass (; ). Most studies reporting CO₂ increases due to cell lysis have observed a short CO₂ pulse lasting up to three days following the rewetting event . Additionally, because we collected our cores during winter, both the soil and microbial communities were subject to freeze–thaw disruption, which could induce thermal shock and amplify cell lysis. In our study, CO₂ released from microbial biomass could partly explain the pulses observed during the first two to three days of incubation, rather than the overall elevated emissions (Fig. ). CO₂ pulses following thaw could also be related to releases of trapped CO₂. Finally, these CO₂ pulses may also be enhanced by physical disturbances due to sequential thawing and, therefore, may not represent natural processes. However, since all the mesocosms were thawed under similar conditions, potential releases of trapped CO₂ or physical disturbances due to setup limitations should be comparable across mesocosms (excluding micro-scale spatial heterogeneity).

The lack of significant differences between water treatments during incubation (Fig. , Table ) is likely due to the overriding effect of water saturation itself, which may have superseded any potential influence of varying water compositions. Additionally, the DOC content of the added water was an order of magnitude lower than that of the palsa and fen soils (9.68; Table S3), thereby reducing the impact of waterborne OM on CO₂ emissions. reported a 60 % decrease in CO₂ production in the permafrost layer under the non-filtered treatment compared to sterile water over a 384 d incubation, likely due to the establishment of anaerobic conditions and the absence of vertical transport. In contrast, our results consistently show approximately twofold higher CO₂ emissions under flooded conditions. We attribute this primarily to vertical substrate mobilization, residual O₂ within the soil column, or redox shifts driven by increased water content, rather than to microbial or chemical inputs from thermokarst pond water. Nonetheless, methodological factors – such as the lateral mode of water addition and data filtering – may have influenced the observed treatment effects.

Finally, we observed CH₄ uptake ranging from 15 to 34.6 µg-CH₄ m⁻² h⁻¹ (Table , Fig. S15), even under flooded conditions – consistent with findings by , who reported uptake ranging from 0.35 to 2.42 mg-CH₄ m⁻² d⁻¹. This observation aligns with results from smaller incubation studies that reported little to no CH₄ production in the active layer . We attribute the observed CH₄ uptake to methane oxidation within the soil column, likely facilitated by the presence of residual O₂. Under in situ conditions, the active layer tends to remain dry and oxygenated, which inhibits CH₄ production.

Palsas in Fennoscandia experience degradation rates that vary between -1.0 % yr⁻¹ and -1.3 % yr⁻¹ over the period of 1950/1960–2010/2014 . Palsa collapse has been identified as a major driver of permafrost degradation , highlighting the importance of abrupt thaw. While our study focuses on short-term CO₂ responses (three-month incubation) driven by hydrological differences between thaw types, these processes also will likely differ in their temporal dynamics. Our results suggest that CO₂ emissions from abrupt thaw peak early and then stabilize and/or decrease as anaerobic conditions develop, whereas emissions from gradual thaw may increase more steadily as deeper layers thaw. The rise in CO₂ emissions from the Control during the final 10 d of incubation may reflect this slower process (Table ). Over longer timescales, palsas undergoing thaw (both abrupt and gradual) transition from initial CO₂ sources to net sinks as vegetation re-establishes and ecosystem productivity increases, a process that may take several decades . This process is difficult to capture accurately under field conditions, highlighting the importance of continuous, long-term monitoring to assess CO₂ emissions during organic matter degradation – particularly before vegetation shifts and the onset of CH₄ emissions.

C mineralization following permafrost thaw from peat partly depends on the stage of peat decomposition . Our FTIR results, along with soil parameters measured prior to incubation, indicate a higher potential for OM decomposition in the permafrost layer (>60 cm) compared to the active layer (0–60 cm), as evidenced by higher pH values and reductions in CH and C=O peaks (Figs. , a). According to the post-incubation FTIR spectra and DI ratios, the highest decomposition occurred at the permafrost table (60 cm) under both flooded and dry conditions (Fig. b). This enhanced decomposition at the permafrost table can be attributed to the presence of active microbial communities – likely translocated from the active layer during thaw – and the availability of relatively undecomposed peat. In contrast, little to no peat decomposition occurred deeper within the permafrost layer (60–100 cm; Fig. b), despite the observed increase in CO₂ emissions under control conditions (Table ). Furthermore, we observed minimal differences in the stage of peat decomposition between the deeper permafrost layer and the equivalent layer in the fen (60–100 cm; Figs. , a). This suggests that both the short-term thaw simulated in this study and 25 years of permafrost-free conditions have not significantly altered OM quality in the deeper permafrost layers (>60 cm). These findings are consistent with observations from other permafrost regions, such as in Canada . Thus, we conclude that peat degradation during permafrost thaw is primarily concentrated at the permafrost table, whereas in deeper layers, neither thaw processes nor long-term permafrost loss have notably affected peat quality.

estimated that the fen site in Peera lost approximately 15 kg C m⁻² since thaw, without indicating any significant increase in peat decomposition at the fen site (Fig. a; ). This finding is consistent with , who measured a net C loss of about 19.3 kg C m⁻² since thaw, with no observable changes in peat quality. Several studies have reported lower CO₂ emissions from degrading permafrost under waterlogged conditions , which could explain the minimal change in peat quality observed in the fen. This is in line with our observation of decreased CO₂ emissions per volume of thawed soil in the abrupt thaw simulation (flooded) and higher emissions in the gradual thaw simulation (Control) during the final thaw step (0–100 cm; Table ). The greater contribution from deeper layers under control conditions may result from increased gas diffusivity in the drier soil column. However, without ¹⁴C analysis, it remains unclear whether emissions originate solely from the 80–100 cm layer or if the 60–80 cm layer contributed later due to lower diffusivity. Under flooded conditions, water saturation limits oxygen availability and slows gas diffusion , leading to anaerobic conditions within the permafrost layer. Although microbial decomposition still occurs, it shifts to less efficient anaerobic pathways that produce CO₂ (along with other gases), which then accumulates due to restricted diffusion through the saturated soil.

In the fen mesocosms, CO₂ emissions increased with depth despite flooded conditions (Table , Fig. ), with the highest DIs observed at the bottom of the cores (Fig. b). While surface vegetation was removed during incubation, the root systems were retained. The presence of deep-rooted aerenchymous plants, such as sedges, may explain the increased CO₂ emissions in the fen. Aerenchyma tissue enhances the transport of O₂, CO₂, and CH₄ in waterlogged soils . Although aerenchyma are best known for enhancing CH₄ transport and reducing CH₄ oxidation, they can also provide pathways for CO₂ movement. Because CO₂ diffuses roughly 10 000 times faster in air than in water , gas exchange through aerenchyma can substantially influence CO₂ fluxes. In this process, O₂ is transported downward to the roots, while CO₂ produced by microbial activity moves upward and is fixed in the shoots via photosynthesis . In the absence of photosynthesis, roots may have acted as pathways for CO₂ from deeper layers. However, the artificial sequential thawing process may have amplified these emissions, leading to potential overestimation of CO₂ fluxes.

Finally, the degradation of dead roots could provide additional substrate, contributing to the higher DI observed at the bottom layer (Fig. b). This process may also increase soil porosity, facilitating gas exchange to the surface. This highlights the potential role of vegetation change since thaw in CO₂ transport, particularly when there is no active vegetation. While several studies have measured and modeled increased CO₂ emissions in northern fens during the non-growing and shoulder seasons , few have accounted for the influence of vegetation type and root structure on CO₂ dynamics. This represents a significant gap in existing research and models, underscoring the need for further investigation in this area.