The ongoing lateral expansion of peatlands in Finland

Peatlands are the most dense terrestrial carbon stock and since the last glacial epoch northern peatlands have accumulated between 400 and 1000 Gt of carbon. Although the horizontal development history of the peatlands during the Holocene has been previously researched, these studies have overlooked the current peatland margins. This has led to a long‐standing view that the lateral expansion of the peatlands has halted or significantly slowed down. However, no concentrated effort focusing on the recent development of the peatland margins has been conducted. To fulfil this knowledge gap, we studied the development of peatland margins in five Finnish peatlands. In addition, we studied the effect of peatland subsoil characteristics and past forest fires on the peatland expansion. We sampled 15 transects with a total of 47 peat cores utilizing 14C radiocarbon dating on the basal layers of these peat cores. Our results show that the Northern peatlands are still expanding with four of our study sites having recent, post‐1950's basal ages in the peatland margins. In addition, the rate of peatland lateral expansion has increased during the last 1500 years in our study sites, challenging the current knowledge of the recent peatland expansion dynamics. We recorded lateral expansion rates of 0.1–6.4 cm/year from the sites studied. The rate of lateral expansion was restricted by local characteristics, especially the steepness of subsoil (p = .0108). Forest fires likely played an important role as the trigger for lateral expansion in southern study sites with large number of charcoal found at the basal layer of the peat cores. Depending on the scope of this recent lateral expansion across the vast northern peatlands, the effect on the carbon balance could be significant and should be taken into account when estimating the development of carbon pools in these crucial ecosystems.

the horizontal development history of the peatlands during the Holocene has been previously researched, these studies have overlooked the current peatland margins.
This has led to a long-standing view that the lateral expansion of the peatlands has halted or significantly slowed down.However, no concentrated effort focusing on the recent development of the peatland margins has been conducted.To fulfil this knowledge gap, we studied the development of peatland margins in five Finnish peatlands.
In addition, we studied the effect of peatland subsoil characteristics and past forest fires on the peatland expansion.We sampled 15 transects with a total of 47 peat cores utilizing 14 C radiocarbon dating on the basal layers of these peat cores.Our results show that the Northern peatlands are still expanding with four of our study sites having recent, post-1950's basal ages in the peatland margins.In addition, the rate of peatland lateral expansion has increased during the last 1500 years in our study sites, challenging the current knowledge of the recent peatland expansion dynamics.We recorded lateral expansion rates of 0.1-6.4cm/year from the sites studied.The rate of lateral expansion was restricted by local characteristics, especially the steepness of subsoil (p = .0108).Forest fires likely played an important role as the trigger for lateral expansion in southern study sites with large number of charcoal found at the basal layer of the peat cores.Depending on the scope of this recent lateral expansion across the vast northern peatlands, the effect on the carbon balance could be significant and should be taken into account when estimating the development of carbon pools in these crucial ecosystems.

K E Y W O R D S
carbon, environmental change, lateral expansion, mires, peatland areas, peatlands, wildfires three times more than carbon stored in the living biomass and soils of forests, making peatland ecosystems the most important carbon storage found in Finland (Liski et al., 2006;Liski & Westman, 1997).
The vertical peat growth of the peatlands in Finland has been widely studied since the early 20th century (Seppä, 2002 and references therein).However, only scant attention has been given to the lateral changes in the Finnish peatland extent during the Holocene (Korhola, 1994(Korhola, , 1996;;Mäkilä, 1997;Mäkilä & Moisanen, 2007;Mathijssen et al., 2017;Weckström et al., 2010), although the lateral expansion is the most important process in new peatland area formation (Ruppel et al., 2013).Even in these studies the ongoing development of the forest or mineral soil restricted peatland margins has been left for lesser consideration.Due to the importance of peatland margins in landscape dynamics, for example, regulating downstream hydrological and biogeochemical patterns, microclimate, local biodiversity and ecosystem carbon balance, knowledge on these transitional zones is sorely needed.
After the peat initiation, which can take place in a single locus or several loci, the area of peatlands is controlled by the lateral expansion of peatland margins.The primary requirement for the lateral expansion-as generally for peat formation-is a slow decomposition of biological materials, which takes place in waterlogged, anoxic conditions resulting in increased biomass inputs to the soil (Clymo, 1984).Such conditions are created mainly by local environmental factors and/or autogenic development of peatland (Anderson et al., 2003;Bauer et al., 2003;Foster & Wright, 1990;Korhola, 1994;Loisel et al., 2013).On flat terrains and surfaces, in the absence of physiographical constraints, the lateral expansion is commonly fast, with peatland fronts progressing even several metres per year (Foster et al., 1988;Korhola, 1994;Mäkilä, 1997;Peregon et al., 2009).Similarly, rapid expansion of peatlands can occur in areas with small-particle size subsoil such as clay or silt that have a large water-holding capacity and are thus permanently inundated (Foster et al., 1988;Kleinen et al., 2012;Korhola, 1994) while pedogenic processes that decrease soil permeability may lead to increased water tables in peatland margins (Rydin & Jeglum, 2013).Waterlogged conditions in peatland margins may also be caused by disturbances, such as a forest fire or clear-cutting, in adjacent uphill areas that reduce evapotranspiration due to loss of vegetation (Bauer et al., 2003;Ikonen, 1993;Korhola, 1996;Novenko et al., 2021;Schaffhauser et al., 2017;Simard et al., 2007;Tallis, 1991).In addition, rise of water tables in peatland margins may also be created by autogenic process of peat accumulation as the surface of a typical raised bog rises above the surrounding mineral soils and the runoff from the peatland crest is directed to the peatland margins (Anderson et al., 2003;Foster & King, 1984;Korhola, 1996;Rydin & Jeglum, 2013).
Climate conditions may also have a considerable control on lateral expansion of peatlands.During the warm and dry period known as the 'Holocene thermal maximum' (HTM) between ca.8000 and 5000 BP (Before Present, where present is 1950 CE), the lateral expansion of peatlands was greatly reduced (Korhola, 1996;Korhola et al., 2010).On the other hand, the intensity of northern peatland lateral expansion increased as the Holocene climate shifted towards cooler and more humid conditions ca. 5000 BP (so-called 'neoglacial cooling'), supporting the link between large-scale climate phases and the lateral expansion (Ruppel et al., 2013).Also, more localized climate effects are known to promote lateral expansion (Korhola, 1995;Turunen & Turunen, 2003).Thus, these local and global processes either individually or together alter the hydrology on areas surrounding peatlands and may lead to a lateral peat expansion (Foster & Fritz, 1987;Korhola, 1994).
In addition to large changes wrought upon the landscape, lateral expansion of peatlands may have a significant effect on the global carbon cycle.For example, during the late Holocene ca. 5 ka onwards, atmospheric methane concentrations started to increase markedly.In some studies, this has been linked to expansion of large areas of wet, minerotrophic fens on northern hemisphere during a humid late Holocene climate phase (Korhola et al., 2010;Ruppel et al., 2013;Seppä et al., 2009).In aapa mires found typically in Northern Finland, peatland margins are drier than the central areas and they support bog-type vegetation (Foster & King, 1984;Laitinen et al., 2007;Ruuhijärvi, 1960).Contradictorily, the margins of the raised bogs are generally wet and characterized by vegetation typical for fens (Aartolahti, 1965).The drier bog-type margins in aapa mires act as a carbon sink while the wet margins of raised mires are a carbon source as methane (CH 4 ) is produced by decomposition in anoxic conditions (Heiskanen et al., 2021;Juutinen et al., 2013;Nykänen et al., 1998).Therefore, the initial climate forcing of the lateral expansion of peatlands is likely opposite in aapa mire and raised mire systems.However, due to the heterogeneous and individual nature of these complex ecosystems, the expansion and carbon accumulation dynamics could differ even within the same peatland area.
Prevailing perception has been that most of the areas climatically and environmentally suitable for lateral expansion have already been covered by peatlands in Finland (Ruuhijärvi, 1983) and in other parts of Scandinavia (Sjörs, 1983), as a consequence of which the rate of lateral expansion has dropped significantly or essentially ceased 2000 to 1000 years ago (Korhola, 1994;Korhola et al., 2010;Peregon et al., 2009;Ruppel et al., 2013).Also, although otherwise suitable for peat formation, in some areas the already prevailing forest ecosystems have proved resistant to the paludification (Ehnvall et al., 2023).However, only a handful of studies across the globe have concentrated on the ongoing changes in peatland margins (Le Stum-Boivin et al., 2019;Peregon et al., 2009;Pluchon et al., 2014;Zhao et al., 2014).In Finland, no concentrated effort has been made to study the recent lateral expansion of the peatland margins (Mäkilä, 1997;Mathijssen et al., 2014Mathijssen et al., , 2017)).Although there are large uncertainties in climate change scenarios, it is expected that the predicted increase in temperature and precipitation in northern Europe (Bednar-Friedl et al., 2023) may accelerate lateral expansion of peatlands due to increased productivity and changes in hydrological conditions (Franzén et al., 2012;Gallego-Sala et al., 2018;Weckström et al., 2010).As the lateral expansion may have a large impact on the formation of new carbon pools in the peatlands in Finland and elsewhere in northern areas, it is crucial to provide up to date knowledge of the development trajectories of peatland margins.
In this research, we studied the occurrence and rate of lateral expansion in several peatland margins, using multiple peat basal ages collected along transects of shallow peat profiles from the peatland edges.In addition, we studied the peatland subsoil characteristics and macroscopic charcoal of basal peat layers to examine the contribution of these above-discussed local factors to the lateral expansion process.Our results have wide implications for boreal landscape dynamics and the high-latitude carbon sink capacity since even a low expansion in peatland area has drastic significance for the future potential carbon sink dynamics when multiplied throughout the vast northern peatland region.

| Study sites
We included five study sites in our research based on a preliminary investigation of suitable areas utilizing aerial images, satellite imagery and digital elevation models.We considered the variations in growing season length (north-south axle) and the differences between more maritime-type (western) and continental-type (eastern) climates by inclusion of study sites across different parts of Finland (Figure 1; Table 1).

| Syysjärvi
Our northernmost study site is located in the southern margin of the Sammuttijänkä-Vaijoenjänkä mire protection area (69°18′ N, 27°14′ E; Figure 1).The sampling location was in the western margin of the large fen basin covered in string-flark patterns and open water areas (Figure 2a).Four transects were taken from the eastward sloping peatland margin area.The vegetation of the margin is a mixture of dwarf shrubs, sedges, feather mosses, lichens and Sphagnum patches.

| Salamajärvi
Our westernmost study site is located in Salamajärvi National Park (63°16′ N, 24°48′ E; Figure 1).We collected the study transects from the western and northern margins of small peatland basin north of Lake Koirajärvi (Figure 2c).Compared to our other aapa mire sites, more minerotrophic vegetation such as Picea abies trees, Vaccium myrtillus and Carex species can be found at the studied margin.

| Patvinsuo
This study site is located in Patvinsuo National Park (63°07′ N, 30°44′ E) in the eastern part of Finland (Figure 1).Patvinsuo (PS) is at the southern edge of the aapa mire region in Finland and features of both aapa mires and raised mires can be found in the area (Turunen et al., 2002).Four transects were taken from the western and the eastern edge of a mineral soil forest cutting through the main peatland basin (Figure 2d).The microforms control the vegetation in the sampled mire margin: in hummocks, small pines, dwarf shrubs and ombrotrophic Sphagna can be found, while in lawns, Eriphorum vaginatum, Rhyncospora alba and Sphagnum balticum dominate.In addition, a single transect was taken in north-south direction from a forested peatland ca.500 m north of the other transects.

| Siikaneva
Siikaneva (SN) mire protection area (61°50′ N, 24°10′ E) is located in the middle of the raised mire region (Figure 1).We collected a single transect of peat samples at the north-eastern extreme of the peatland area in a narrow patch of the peatland penetrating into the surrounding mineral soils (Figure 2e).Minerotrophic vegetation was present at the peatland margin, and large quantities of Polytrichum species were mixed with Sphagmum mosses in the ground layer.Of our study sites, SN is most heavily affected by direct anthropogenic influence, with a recently performed clear-cut in the adjacent forest likely affecting the hydrology in the studied margin.

| Field sampling
Prior to collecting the peat cores from our study sites, we estimated the location of the peatland margins of our study sites using aerial images and base maps acquired from the National Land Survey of Finland.The exact location of the peatland margins was closely inspected in the field based on the presence of the peatland vegetation, mostly Sphagnum mosses.We collected transects of peat cores from the peatland margins towards the peatland centre (henceforth abbreviated as T followed by the transect number).These transects were divided into horizontal segments (HSEG tagged by letter: m = segments closest to the peatland margin, i = intermediate segment(s) and p = segment closest to the peatland centre) between the sampled peat cores or a peat core and the specifically defined position of the peatland margin in the case of the HSEG closest to the peatland margin.The length of each HSEG was measured in the field, and location and elevation (m.a.s.l.) of sampling point of each peat core was recorded using a GPS device.A total of 47 peat cores along 15 transects were sampled during summers of 2018 (Ka study site), 2019 (PS, Salamajärvi [SJ] and Syysjärvi [SyJ] study sites) and 2020 (SN study site; Table 1).We took 43 samples using a box corer (7 × 4 × 65 cm diameter) and four samples using a saw in cases where the subsoil was extremely rocky.The core materials were described and classified in the field.Whenever possible, mineral subsoil was also included in the sampling for subsequent, more comprehensive analysis of the subsoil characteristics.We carefully wrapped the samples in plastic to avoid contamination and transported them to the University of Helsinki premises.The samples were stored in a cold room prior to further analysis.In addition, high-resolution mapping of the subterranean topography in PS and SJ study sites was performed using a peat probe and the peatland surface positioning by GPS.

| Laboratory analyses
We cut the peat cores first into 1-cm subsamples.From these subsamples, mineral-to-peat transition for each core was detected F I G U R E 1 Lateral expansion study sites: Syysjärvi, Kaamanen, Salamajärvi, Patvinsuo and Siikaneva.The map also shows location of the main mire complex types in Finland: raised mires in the south and aapa mires further in the north.Map lines delineate study areas and do not necessarily depict accepted national boundaries.

TA B L E 1
Study site coordinates, mean annual precipitation (mm), mean annual temperature (°C) and number of transects and peat cores sampled from each study site.

F I G U R E 2
The study locations within the study sites: (a) Syysjärvi, (b), Kaamanen, (c), Salamajärvi, (d), Patvinsuo and (e) Siikaneva.The location of collected transects are marked with a black square with white stripes.
quantitatively using the sediment organic matter (OM) content based on the loss on ignition (LOI) method as recommended by Quik et al. (2022).We considered the basal layer to be the first layer from bottom towards surface with LOI ≥70% (Korhola, 1994).Four peat cores were taken from the top of mineral subsoil consisting mostly of boulders.For these samples, the layer found directly on the top of the boulders, for example, the bottom sampled layer, was considered the basal peat.We also determined the bulk density (g/cm 3 ) for basal layers using known volume and weight measurements.
To determinate the age of the bottom peat layers of our transects, we applied AMS radiocarbon ( 14 C) determinations to date the basal subsample with thickness of 1 cm for each studied core.We examined the basal subsamples with stereomicroscope and chose plant macrofossils and/or charcoal from each subsample to be dated (Table 2), as plant macrofossils have been shown to yield the most accurate age in the mineral-to-peat transition and are therefore recommendable to use for 14 C dating basal peat (Quik et al., 2022).In five samples, the highly decomposed peat made this impossible, and instead, we used bulk peat with roots removed.We sent the chosen materials to two 14 C AMS dating laboratories.Twenty-five of the samples were dated in the Laboratory of Chronology by the Finnish Museum of Natural History (LUOMUS) and 22 of the samples in Poznan Radiocarbon laboratory (Poznan, Poland; Table 2).We used Oxcal version 4.4 (Ramsey, 2009) to calibrate our basal peat samples into calendar ages.
The 14 C basal ages acquired from the dating laboratories were calibrated using INTCAL 20 NH1 calibration curve (Reimer et al., 2020) while the modern dates (pMC (%) modern carbon) were calibrated using Bomb21NH1 calibration curve (Hua et al., 2022).
To analyse the mineral subsoil fractions of the study sites, we used mineral material collected alongside the peat cores directly underneath the first basal layer of each sampling point.Out of 47 sample cores, 39 included sufficient mineral material for mineral fraction analysis.Homogenized subsoil samples of 10 cm 3 were first dried in the oven for 12 h in 105°C to measure the water content (%) and dry bulk density (g/cm 3 ).Then, the OM was removed using LOI method (Heiri et al., 2001).Prior to the laser diffraction analysis, the pretreated samples were sieved using 0.6 mm mesh as the inclusion of coarse material may reduce the accuracy of laser diffraction analysis.
The material removed by sieving was measured to acquire mineral soil fraction for coarse material.Coulter LS230 laser diffraction analyser was used to acquire mineral soil fractions ≤0.6 mm.The resulted mineral subsoil fraction was then categorized into clay (<0.002 mm), silt (0.002-0.05 mm) and sand components (0.05-2 mm).
We also studied the presence of wildfires during the initiation period of the basal peat layers by quantifying the number of charcoal particles in the peat basal layers. 1 cm 3 subsamples of the basal peat layer from peat cores were studied with stereomicroscope.
Two charcoal size classes were used: small charcoal particles with a diameter <1 mm and large charcoal with diameter ≥1 mm with the assumption that microcharcoal records ≥1 mm represent local scale fires either in situ or within a few 100 m from the study site (Carcaillet et al., 2001;Sim et al., 2023).Up to 100 charcoal pieces for each charcoal size class were counted.In addition, we used non-parametric Mann-Whitney U-test using base R software ver.
4.2.2 (R Core Team, 2022) to test for a statistical difference between the amount of large-and small-sized charcoal in the northern sites of SyJ and Ka and the southern sites of SJ, PS and SN (data: Juselius-Rajamäki et al., 2023a).SJT2S i and SJT2S p were excluded from the analysis due to lack of peat material.
Finally, we studied the overall composition of the peat samples to detect the peat types across the sampled peat cores.We used stereomicroscope to detect macroscopic plant remains in the peat samples and categorized the peat as either woody/Carex type or Sphagnum type.In many international, especially geological, classifications the peat depth requirement for a peatland varies but is commonly considered to be ≥30 cm (Ågren et al., 2022).However, as these requirements are seldom fulfilled in the thin peat layers of peatland margins and other "cryptic wetlands", we used the presence of the peatland vegetation, in particular Sphagnum mosses, as further evidence of our study locations to be classified as a peatland.In such case, the shallower threshold of organic layer thickness to define a peatland included more of mineral soil wetland that are dominated by living peat-forming plants and often have a substantial content of OM (≥70%) within their surface layers.

| Peatland lateral expansion rate and peat vertical increment calculations
To study the peatland area expansion of our study sites, we first calculated the peatland lateral expansion rate (cm/year) for each HSEG by dividing the distance between the adjacent cores by the difference in the basal ages of the cores.For the current ultimate peatland margin point, sampling year was used.The following Equation (1) was used to acquire the expansion rates for our horizontal segments, that is, over the dated coring points.
To calculate the mineral subsoil slope in our study HSEGs, we used the depth of the basal layer deducted from the surface elevation (m.a.s.l.) of each sampling location and the length of the HSEG using the following Equation ( 2): where θ = slope (°) of the HSEG, ∆y = the vertical change along the HSEG and ∆x = length of the HSEG.
To study the differences in the lateral expansion rates between HSEGs from the peatland margin to the first dated basal age (e.g. the first HSEG) and HSEGs with only dated basal ages we used non-parametric Mann-Whitney U-test using base R software ver.
4.2.2 (R Core Team, 2022).In addition, to compare our results to the Holocene lateral expansion rates from other peatlands in Finland, we used expansion rates given in the study by Korhola (1994).
Furthermore, we calculated slopes (°) from the figures presented (1) lateral expansion rate = distance between cores basal age a − basal age b . ( TA B L E 2 Peat core information and radiocarbon dates.Sample location shows the location of each peat core on their corresponding transect: the margin between the peatland and the forest (Margin), the core or cores taken further away from mineral soil (Intermediate) and the core taken at the end of the transect furthest on the peatland (Peatland).Basal depth (cm) is the depth of the sample used on the 14 C dating based on the loss on ignition analysis.Depths marked with an asterisk (*) have been sampled from rock surface instead.Sample description shows material used in radiocarbon dating. 14 in the same study using Gnu Image Manipulator Program version 2.10.10 and divided both our lateral expansion rates and those presented by Korhola (1994) into a three slope groups (<1°, 1-2° and >2°) for further investigation on the effects of the mineral subsoil slope on peatland expansion (Figure 11).
To study the trend in lateral expansion of peatlands across our study sites, we calculated average peatland expansion rate (cm/year) using data between the dated peat cores in 10-year intervals to examine the trend in the peatland expansion rates.Locally estimated scatterplot smoothing was then applied to these average values using ggplot2 package version 3.4.1 (Wickham, 2016) in R software ver.4.2.2 (R Core Team, 2022).
Finally, to study the connection between the vertical growth and lateral expansion of the peatlands, we performed a Spearman's rank correlation analysis suitable for non-parametric data in R software ver.4.2.2 (R Core Team, 2022) using cor.testfunction.We first calculated the annual vertical growth of the studied peat cores by dividing the depth of the basal layer by the basal age.Prior to correlation analysis, five outliers with exceptionally high horizontal and/or vertical growth were detected using boxplots and removed (SyJT1Sm, SyJT2Sm, SyJT3Sm, SJT2Sm and SNT1Sm, and corresponding segments).For each study segment, we tested the correlation between the vertical increment (mm/year) of the peat core at the start of the segment against the lateral expansion rate of the corresponding segment (data: Juselius-Rajamäki et al., 2023b).

| Statistical analyses
In previous studies (Almquist-Jacobson & Foster, 1995;Korhola, 1994Korhola, , 1996;;Loisel et al., 2013;Zhao et al., 2014), a clear relationship between mineral subsoil slope and the rate of the lateral expansion of peatlands has been found.It has also been proposed that lateral expansion of peatlands is faster on the fine fraction mineral subsoils that are characterized by high moisture content (Almquist-Jacobson & Foster, 1995;Ehnvall et al., 2023;Ivanov, 1981;Korhola, 1996).However, to our knowledge this has not previously been analytically tested.In our study, we combined clay and silt content to a fine fractions (%) variable to analyse this effect.An average of the adjacent peat core fine fractions (%) was used to represent approximation of the subsoil fractions along the segments.In addition, we used the latitude of each of the study sites as a proxy for differences in climatic conditions.
In this study, we used a gamma generalized linear model (GLM) with a log link function to test whether the covariates subsoil slope, fine mineral subsoil fractions and latitude affected the lateral expansion rates of our study segments between dated basal peat samples (Equation 3; data: Juselius-Rajamäki et al., 2023c).The gamma distribution was chosen due to its suitability for right-skewed positive continuous data, while the log link ensures positive fitted values.
Total number of segments studied was 27 and we used a significance level of .05 for our statistical test.

| Modelling peatland characteristics to track recent changes in peatland area
To study how the recent lateral expansion has affected the overall peatland area on our study regions, we modelled the development and 2000-1000 BP.These study sites were chosen as the high-resolution landscape topography determined from these sites allowed us to include the effect of the basal terrain slope for the areal inventory.To calculate the lateral expansion rate across each subterranean topographical transects, we used the following inversed Equation ( 4) of log link function from our GLM: where β 0 = intercept, β 1 = slope (°), β 2 = fine mineral subsoil fraction (%) and β 3 = latitude of the study site.For modelling purposes, we used the average fine mineral subsoil fraction of the corresponding study site as a constant value for β 2 .We then calculated the location of the peatland margin at the start and end of the chosen time intervals for our subsoil topographical transects and used these points to form the peatland area for PS and SJ study sites as 1000 BP and 2000 BP.Finally, we used ArcGis Pro GIS software version 3.1 (ESRI, 2023) to calculate the current peatland area, the peatland area at 1000 BP and at 2000 BP and the increase in the peatland area at each time interval.

| Basal peat ages
Our basal dating results show recent lateral expansion in the margins of all our study sites (Table 2).In general, the age and the thickness of the peat layers decreased towards the peatland margin (Figure 3a-o).The stratigraphy of the peat cores showed that the lateral expansion had occurred over forest soils with abundance of woody material in the basal peat samples.In some cores, high amount of Sphagnum remains were found directly on the top of the mineral subsoil while in majority of the cores the Sphagna appeared only in upper layers of the peat.In seven study transects, young basal peats initiated since 1950 CE (BP 0) were found in the peat cores closest to the peatland margin (Figure 3a-c,h,k,m,o).In the margin cores from KaT2, SJT3 and PST1, basal ages of 110 (20-270) cal BP, 180 (0-300) cal BP and 120 (0-280) cal BP were recorded respectively (Figure 3f,i,j).The margins of five transects showed older basal dates from 460 (320-510) to 3180 (3070-3330) cal BP (Figure 3d,e,g,l,n).In 12 of 15 study transects the peatland margins had expanded progressively and rather uniformly.However, in the KaT1 and KaT2 as well as in the SJT1, the oldest basal ages were found along the middle of the transects.The oldest peat in the KaT1 was found in the middle of slope rather than in the lower parts of the transect as one would expect (Figure 3e).In KaT2 the peat had started to form in several loci both in the slope and in a small recess further along the transect (Figure 3f) while in the SJT1 the peat had started to form in a deeper recess between higher mineral ridges (Figure 3g).
In addition, much higher values were detected between the dated basal peat samples closest to the peatland margin and the peatland edge (Figure 5).The lateral expansion rate for these values ranged from 0.1 cm/year (KaT1 HSEG m ) to 80 cm/year (SyJT2 HSEG m ).The median for all these points was 2 cm/year with an IQR of 1.3-17.2.These values were significantly higher compared to the expansion rates calculated between other HSEGs with the dated basal peat samples (Wilcoxon rank sum test, p < .001).
We found only weak positive correlation between the lateral expansion rates (cm/year) of HSEGs and the vertical peat increment (cm/ year) on our study sites with R(df = 40) = 0.29 and p > .05(Figure S1).
Since the start of our data series, ~7500 BP the average lateral expansion rate increased slowly towards ca.4000 BP (Figure 6).For the next ~1000 years, the rates remained constant at 0.75 cm/year.
After this, the average lateral rates decreased slightly with a slowest rate reached ca.2000 BP.Towards the present, the lateral expansion rates started to increase again and highest point in our dataset is reached at the end of our time series with a rate of 1 cm/year.

| Mineral subsoil characteristics
The underlying terranean slopes between 0.2° (PST5 HSEG p ) and 11° (SJT2 HSEG m ; Figure 7) were found in our study sites.In PST3 the peatland margin had spread downhill after crossing a higher mineral ridge, giving a slope value of −0.9° for HSEG p.The median of subsoil slopes in our study sites varied from 1.3° (IQR = 0.8°-3.5°) in PS to 4.8° (IQR = 4.0°-7.4°) in SJ.Compared to the other sites, subsoil slopes were markedly steeper in the SJ.

| Charcoal particles
Small charcoal particles were more common than large charcoal particles in the basal layers of our study sites.In total, small particles (4) y = e ( 0 + 1 x 1 + 2 x 2 + 3 x 3 ) , were found in 29 of 45 studied basal peat samples while large particles were found in 24 samples (Figure 9).For the small charcoal particles, the number of cores with either very low or very high count of particles were distributed rather evenly, while cores with medium number of small charcoal particles were uncommon.For large charcoal particles, a median of 2 (IQR 0-25) was recorded.We detected a significant difference between the northern sites of SyJ and Ka compared to the southern study sites of SJ, PS and SN, with higher presence for both small and large charcoal particles found in the basal layers in the South.In the more southern sites, only 2 basal peat samples out of 25 showed neither large nor small charcoal particles while in the northern sites no charcoal was detected in 11 The distance from the peatland margin (m) is positive towards the peatland centre and negative away from the peatland.The elevation is presented as metres above sea level (m.a.s.l.).The vegetation is presented to give a rough impression of the real-life conditions in the study transects and is not in a true scale.In addition, peat types (Sphagnum and Woody/Carex type peat) and mineral subsoil type are presented.The ratio between x and y axes varies between the subfigures. of 20 samples.In the more southern sites, small charcoal particles were missing from three samples and in the North no small charcoal particles were found from 13 samples.Four basal samples without large charcoal particles were found in the southern sites while corresponding number in the northern sites was 17.The non-parametric Mann-Whitney U-test showed statistically significant difference (p < .001)for both charcoal particle size classes between the southern and northern sites.

| Statistical analyses
The results from our statistical test showed that steeper slopes slow the lateral expansion rate with an increase of 1 unit of slope (°) decreasing the lateral expansion rate by 22.7% if other variables remain constant (Table 3).This effect was statistically significant (p = .0108).Increase in fine fractions by unit of 1% accelerates the lateral expansion rate by 2.4% and increase in latitude by unit of 1° slows lateral expansion rate by 10.7%.However, no statistical significance was detected for the fine fraction content (%; p = .075)and latitude (p = .143)at significance level α = .05.McFadden's pseudo r 2 of .2suggests a good fit for the model (Lee, 2013;Louviere et al., 2000).
The root mean square error between the observed and the predicted lateral expansion rates from our dataset is 1.19.With expansion rates higher than 1 cm/year the predicted values are clearly lower than the observed values, while with expansion rates lower than 1 cm/year the predicted values are slightly higher (Figure 10).

| Implications for the recent changes in peatland area
In SJ sub-basin the peatland area increased from 11.3 to 11.6 ha between 2000 BP and 1000 BP, and from 11.6 to 12 ha between 1000 BP and the present (Figure S2; Table 4).Total increase in SJ area for the whole study period was 6%.The peatland area in PS subbasin increased from 49.5 to 51.6 ha between 2000 BP and 1000 BP (Figure S3; Table 4).In latter study period between 1000 BP and the present the area increased 1.3 ha.The total increase in studied sub-basin area in PS region from 2000 BP to the present was 3.4 ha which equals to increase of 7% in area to its present extent.

| DISCUSS ION
In ecological peatland literature there is a commonly held view that during the recent millennia the expansion of the existing peatlands has considerably halted down or even ceased in Finland and elsewhere in Scandinavia.This is predominantly due to the postulation that peatlands have long since taken all they can and there are simply no more topographically suitable areas left for a new peatland formation.However, this presupposition has never been actually tested using the field data.
Our results presented here are in strong contrast with any sort of expansion halt in the recent peatland expansion.On the basis of the extensive number of basal peat ages from the marginal areas of the peatlands studied, we can convincingly state that the lateral F I G U R E 4 Lateral expansion rates (cm/year) for each study site.The black line shows median for each study site and the black dots represent outlier values.The boxplot shows the data from the first to the third quartile and the whisker shows the whole data range for each study site.

F I G U R E 5
Comparison of lateral expansion rates (cm/year) between the values from peatland margin to the first dated basal peat sample (HSEG m ) and the values of HSEGs between the dated basal peat samples.The black line shows median for each study site and the black dots represent outlier values.The boxplot shows the data from the first to the third quartile and the whisker shows the whole data range for each study site.HSEG, horizontal segment.expansion of peatlands is still an ongoing process in Finland.This lateral expansion is currently occurring at the expense of surrounding forested area, as indicated by bottom peat samples containing large amount of woody remains and soils typical for forests with fungal hyphae and charcoal.Moreover, our finding suggests that instead of halting down, the lateral expansion rates have actually increased towards the present since ca.1500 years ago.The most important factor controlling the lateral expansion rate was the underlying terrain topography showing that steeper slopes significantly reduced the rate of lateral expansion.In the southern sites SJ, PS and SN the charcoal particles were present in almost all basal samples, suggesting that fires are one of the driving forces promoting lateral expansion.However, in the northern sites of SyJ and Ka the charcoal particles were uncommon, suggesting that peatlands are capable of expanding also without the boost of forest fires if the other requirements for peatland expansion, such as topographical setting, underlying less permeable soil type and suitable climate, are fulfilled.Only weak correlation was found between the lateral expansion rates and the vertical peat increment in our study sites.
After initial low rates of lateral expansion since the start of the dataset (ca.7500 BP), the expansion rates started to accelerate around 4500 BP (Figure 6).This is in accordance with the results from previous studies showing that during a warm and dry climate F I G U R E 6 Lateral expansion rates (cm/year) of the horizontal segments between the dated peat cores.The dashed black line shows the average expansion rate (cm/year) of all the study segments calculated in 10-year intervals smoothed by locally estimated scatterplot smoothing method with error margin shown in grey.

F I G U R E 7
Mineral subsoil slope (°) for each of our study sites.The horizontal line inside the boxplots shows median for each study site, and black dots represent outlier values.The boxplots show the data from the first to the third quartile and the whisker shows the whole data range for each study site.phase in Europe between ca.8000 BP and 5000 BP the lateral expansion of the peatlands was slow, while the cold and humid conditions from 5000 BP onwards promoted the peatland expansion (Korhola, 1992(Korhola, , 1994;;Mäkilä, 1997;Quik et al., 2023;Ruppel et al., 2013;Weckström et al., 2010).In our sites, the lateral expansion rates reached their highest values around 3500 BP corresponding an extremely cold and wet climate phase in Fennoscandia as deducted from palaeoclimate records (Seppä et al., 2009;Wastegård, 2022).
Afterwards, when climate changed warmer and drier as suggested by proxy records (Luoto & Nevalainen, 2015;Seppä et al., 2009), the peatland expansion rates in our sites decreased reaching lower values around 2000 BP.
Interestingly, and what seems to be the main finding of our study, ca.1500 BP, the lateral expansion rates have started to increase F I G U R E 8 Mineral subsoil content (%) for clay, silt and sand fractions for each study site.The horizontal line inside the boxplots shows median for each study site, the boxplot shows the data from the first to the third quarter and the whiskers shows the whole data range for each study site.
F I G U R E 9 Charcoal particle counts for the basal peat samples.Two charcoal particle size groups are presented in the figure: large, representing charcoal with diameter equal to or larger than 1 mm, and small, representing charcoal with diameter smaller than 1 mm.Up to 100 charcoal particles were counted from the basal layer of each peat core.Due to lack of sample material, no charcoal counts were performed for the samples SJT2 i and SJT2 p .again, and this trend continues to the present day.As already notified, this finding contradicts the long-held view of peatland expansion halt and also the extensive data compilation by Ruppel et al. (2013), which reported that peatland expansion has slowed down during the last 2000 years throughout the northern parts of Europe and North America.However, it is also stated in the Ruppel et al. (2013) that this decline in the expansion rates may be a bias merely reflecting overall lack of data from marginal peatland areas and accordingly lack of young basal peat ages.Our results now cover this data gap for the first time and are further supported by the sporadic results by Mäkilä (1997) and Mathijssen et al. (2014Mathijssen et al. ( , 2017) ) et al., 2013;Seppä et al., 2009;Wastegård, 2022).According to our data the lateral expansion development closely followed these climate patterns suggesting that the expansion dynamics in a bigger picture was controlled by climate.However, recent studies have shown a current drying trend in the peatlands of the North (Juselius et al., 2022;Zhang et al., 2020) that could have an adverse effect on the peatland expansion.It can also be hypothesized that during dry and warm climate periods, the peatland margins may have actually retreated instead of expanding due to peat oxidation.In addition, severe fires are known to cause the peatland margins to temporarily retreat (Lukenbach et al., 2015).However, after peat oxidation, little or no evidence is commonly preserved in the peat archives, and thus studying this phenomenon requires such data not collected for this study.Also, no evidence of any interruption in the peat growth patterns was observed in any of the studied peatlands.In contrast, they all showed very typical peat sequences with continuous root and moss stem structures throughout the peat cores.Conversely, it has been proposed that in areas where growing season length restricts the growth, warming may also promote Sphagnum habitat suitability and peat development (Jones & Yu, 2010;Loisel et al., 2014;Ma et al., 2022).Thus, further study on the lateral expansion of peatlands F I G U R E 1 0 The observed and the predicted lateral expansion rates (cm/year) using our GLM.across varying climate conditions is needed to better understand the climatic drivers behind the area change of the peatlands.
The actual expansion rates between 0.1 and 6.4 cm/year found in our study during the recent times are low (Figures 4 and 6) compared to some previous studies capturing the whole Holocene.For example, Korhola (1992Korhola ( , 1994) reported lateral expansion rates ranging from 1.7 to 800.9 cm/year from four peatlands in southern Finland.
In some Alaskan and Canadian peatlands, the long-term expansion rates have varied from 0.3 to 71.4 cm/year (Asselin & Payette, 2006;Lacourse et al., 2019;Le Stum-Boivin et al., 2019;Loisel et al., 2013) while in Siberian peatlands values between 1.4 and 791.7 cm/year have been reported (Peregon et al., 2009;Pluchon et al., 2014).As the expansion rates in our study sites were quite moderate compared to these previous studies covering also the comparable climate periods, such as HTM, MCA and LIA, the lateral expansion in the peatlands studied here is likely controlled by additional local environmental factors that have contributed markedly to the final architecture and dynamics of the peatland expansion.
Based on our data, the most important local driving factor is the slope steepness of the underlying subterrain (Table 3).In majority of our study segments, the steepening of the slope slowed the expansion of the peatlands while a gentle slope promoted expansion.This is in accordance with the findings by, for example, Korhola (1994), Almquist-Jacobson and Foster (1995), Mäkilä (1997), Loisel et al. (2013) and Zhao et al. (2014).As the studied peatlands are largely located on steep slope terrains, this can at least partly explain the low lateral expansion rates.We analysed this conclusion further by comparing our results to a previous study from peatlands in southern Finland located in exceptionally flat, clay-covered areas of old sea bottoms (Korhola, 1994).
As can be seen in Figure 11, lateral expansion rates from our study are generally slow in all slope groups.In addition, as these differences occur irrelevant of the time periods, the large-scale climate patterns are likely not the main contributing factor.Although sites studied by Korhola (1994) were located further south, based on the low correlation between latitude and lateral expansion rates (Table 3), climate variations between northern and southern Finland likely had only a small effect on the detected differences.Thus, we may conclude that although subterranean slopes have major control over the peatland expansion, there are also other factors to be considered.
A majority of our basal peat samples were collected from areas with coarse, sandy subsoils and more silty materials were found only at SyJ (Figure 8).Coarse subsoils have high infiltration rate, and they are resistant to waterlogging while clay-rich soils act the opposite way.Accordingly, rapid peatland expansion has been reported on clay soils (Korhola, 1994;Le Stum-Boivin et al., 2019;Mäkilä, 1997).In our case, the effect of the subsoil substrate was small, but nevertheless, on areas with high amount of coarse subsoil fractions, the lateral expansion rates slowed down.Thus, regarding our data, in addition to the steep subterranean slopes, the coarse subsoils likely have resulted in relatively slow peatland expansion rates.However, as our study sites lacked the clay subsoil substrates, more information of the recent lateral expansion on such clay terrains is needed.
Not all lateral expansion can be explained by any one local factor alone.As the lateral expansion is controlled by the waterlogging of the peatland margins, changes adjacent to the peatlands, such as forest fires, may have a drastic effect on soil moisture content and hence the peatland expansion.In three southern sites SJ, PS and SN basal peat samples typically had large number of macroscopic charcoal, while in the northern sites the amount of charcoal was much lower (Figure 9).The forest fire frequency varies considerably between northern and southern Finland (Aakala, 2018;Pitkänen, 2000;Pitkänen et al., 2002).The difference is related to the availability of dry biomass fuel, length of the snow cover and overall moisture conditions (Larjavaara et al., 2004).In addition, in southern Finland long-term and more intense human influence has had a large effect on forest fire frequency (Wallenius, 2011;Wallenius et al., 2010).
Low severity fires may benefit peatland expansion by killing trees and reducing evapotranspiration and by promoting light availability creating favourable conditions for Sphagnum moss colonization (Le Stum-Boivin et al., 2019;Novenko et al., 2021).For instance, in PS area lateral expansion has been promoted by recent fires (Turunen et al., 1999(Turunen et al., , 2002)).On the other hand, severe fires may destroy peatland vegetation and peat layers impeding or delaying lateral peatland expansion (Lukenbach et al., 2015;Schaffhauser et al., 2017;Simard et al., 2007).It has been proposed that forest fires are likely to become more frequent (Senande-Rivera et al., 2022), and depending on their severity, they may either promote or reduce the peatland expansion.
Our GLM performed satisfactorily in predicting lateral expansion rates under circumstances where the expansion rates were primarily controlled by local factors.However, as we did not include a temporal dimension in our model, potential promoting or hindering factor by climate variation during the Holocene is missing in the prediction outcome.This may at least partly explain the underestimation of predicted values in cases where the observed lateral expansion rates were higher than 1 cm/year.By examining the individual data in detail (Figure 6), such high expansion rates occurred mostly either during the last 1000 years or between 2500 and 5000 BP; both periods being wet and cool (Hanhijärvi et al., 2013;Seppä et al., 2009;Wastegård, 2022).
Hence, the model likely underestimates the lateral expansion rates, as it does not include the effect of the climate fluctuations.Thus, larger dataset with observations from a wider range of local conditions as well as the inclusion of temporal climate factor would be needed to make the model more suitable for predicting peatland lateral expansion under various environmental and climate conditions.While this recent peatland expansion may be considered modest by the first insight it is still a considerable amount of new peat formation when multiplied over the vast northern peatland region.
Based on our recent experience on the study of the peatland margins, we propose a careful topographical investigation of the mineral subsoils in chosen study area prior to actual sampling.Due to the possibility of peat initiation at multiple loci in deeper recesses and subsequent coalescence to a single peatland mass, even a small difference in sampling location could have a significant effect on the acquired basal peat ages.Also, as our results show, transects even a few metres apart have large variation in the basal ages and lateral expansion rates.Thus, we suggest multiple adjacent transects to ensure that collected samples successfully represent the development of the peatland margin in the chosen study location.In addition, although many studies handling lateral expansion dynamics have highlighted the importance of mineral subsoil substrate in a general level, the quantitative approach used in our study is usually not applied, challenging detailed data comparison.As the sampling and analyses of the subsoils underlying shallow marginal peats is relatively easy, such practice should be incorporated into future studies when lateral expansion is investigated.

| CON CLUS ION
We studied recent lateral expansion of peatland margins in five Finnish boreal peatlands.Our study shows that lateral expansion is still an ongoing process across Finland.By sampling several parallel transects from the same peatland margins, we were able to discern large variations in lateral expansion rates occurring even in a relatively small area.In addition, contrary to the current understanding of peatland expansion history, our sites show an increasing trend on the lateral expansion rates during the last 1500 years, which continues to the present day.Lateral expansion occurred also in locations with rather steep subterranean slopes and coarse-grain subsoils.
However, the rates of lateral expansion in our study sites were slow compared to studies conducted in peatlands located in flat terrain with fine-grained subsoils, showing the importance of the local factors on the peatland expansion.
We still lack knowledge of recent development of the northern peatland margins across different geographical regions and local settings.Therefore, we propose further large-scale studies on the lateral expansion of peatlands to reveal the full scope of this phenomenon and the factors controlling it.Nevertheless, it can be speculated that if this expansion pattern was occurring throughout the northern peatlands, this would have a large effect on the carbon sink and storage balance and should be taken into account when estimating the carbon pool of northern peatlands.

F
The transect profiles (a-o) of our study sites.The basal ages of each sample core are shown in the figure as cal BP years.
altogether demonstrating accelerated peatland expansion in Finland during the recent few millennia.In addition, apart from the warm and dry Medieval Climate Anomaly (MCA; ca.900-1200 CE), the last 1500 years have been particularly favourable for peatland expansion with cool and humid climate phases, the Dark Ages Cold Period (600-900 CE) and the Little Ice Age (LIA; 1500-1850 CE; Hanhijärvi

TA B L E 4 F
Increase in peatland sub-basin area since 2000 cal year Before Present (BP) for Salamajärvi and Patvinsuo study sites.The calculations are based on the gamma GLM.I G U R E 11 Comparison of lateral expansion rates in this study and study by Korhola (1994) from peatlands in Southern Finland in three different slope angle groups: Slope angle (°) <1°, slope angle (°) 1-2° and slope angle (°) > 2°.Please note that y-axis values vary between the slope groups.
SJ and PS study sites contained detailed information about their basin topography for studied sub-basins.We utilized GLM to study peatland expansion in these two study sites since 2000 BP using a sub-basin level approach.The variations in the development rate of the peatland area were mostly constructed based on the changes in subterranean topography, as the effects of fine fraction of subsoil and latitude were calculated based on constant values of each study site.In SJ, the steep subterranean slopes have limited the lateral expansion of the peatland in the northeast and southwest part of the peatland (FigureS2), and the expansion has occurred on the shallower northwest and southeast slopes.Based on our reconstruction, 94% of the sub-basin area was covered by peat by 2000 BP.In PS, the expansion was slightly more rapid, and the peatland margins seem to have advanced nearly linearly at all forest-peatland borders due to relatively evenly rising subterranean slopes (FigureS3).Nevertheless, our results show that 93% of the peatland in the sub-basin had formed already prior to the 2000 BP.Although our model likely underestimates the lateral expansion rates, it appears that only around one tenth of the peatland area in sub-basins of SJ and PS study sites have formed since 2000 BP.

Study site Coordinates Mean annual Number of Latitude Longitude Precipitation, mm Temperature, °C Transects Cores
C age (Before Present[BP]) is an uncalibrated age of the sample with standard errors (±) and pMC represents modern carbon % with standard errors (±).Age (cal.BP) represents median calibrated sample age with 95.4% probability range in parentheses.

TA B L E 3
Results of the gamma GLM statistical test (degrees of freedom = 23).