Depositional history of peatland pines (Pinus sylvestris L.) in NW Enontekiö, Finnish Lapland: implications for Middle Holocene drought and temperature fluctuations

High altitude and latitude findings of subfossil peatland pine trees were unearthed from the region of NW Finnish Lapland and dated by 14C and tree‐ring methods. The depositional history of the trees illustrated two distinct peatland pine phases dated to Middle Holocene intervals 4900–4400 and 4100–3400 cal. a BC. It seems evident that both thermal and hydroclimatic fluctuations have played roles of varying importance in the establishment of this pine population and its demise. The presence of these pines, from a site ~60 km north of the coniferous timberline and conditions ~1 °C and 100 degree‐days colder than those at the present‐day timberline, concurs with previous studies demonstrating the association between the high‐latitude summer‐temperature cooling and circumpolar timberline retreat since the Middle Holocene due to Milankovitch forcing. On the other hand, the peatland pine recruitment was made possible by drier than present surface conditions during the previously reconstructed Middle Holocene drought anomaly (Hyvärinen‐Alhonen event). Our data suggest this event was not continuous but reached its two‐phase climax during the peatland pine phases, with an interruption of several centuries with moister surface conditions between 4400 and 4100 cal. a BC. The findings highlight the sensitivity of well‐dated peatland tree assemblages in terms of recording past climatic evolution and events and the need for new collections from north and south Fennoscandia and the Baltic region, for more detailed analyses over extended time intervals and regions.

Subfossil remains of trees found in treeless peatlands intrigued the minds of early naturalists from the 18th to 19th centuries (Birks 2008). In Denmark, the puzzle was even more striking as pine had gone extinct as a native tree species. It was there that Quaternary palaeoclimatic research in Fennoscandia began. The Danes Christian Theodor Vaupell and Johannes Japetus Smith Steenstrup proposed theories explaining the occurrence of buried pine trees by long-term fluctuations in climatic and edaphic conditions, consisting of changes in moisture, temperature and sunlight (Birks 2008;Birks & Sepp€ a 2010;Nielsen & Helama 2012). Today, analyses of climatic changes have become a backbone of Quaternary research, but the subfossil peatland pine findings remain an enduring topic of much interest for palaeoclimatic and palaeoecological studies. Subfossil pinewood samples are routinely dated using 14 C and/or dendrochronological methods and their assemblages explored for depositional histories indicative of long-term (low-frequency) climatic variability and peatland development (Leuschner et al. 2007;Eckstein et al. 2009;Moir et al. 2010;Edvardsson et al. 2016bEdvardsson et al. , 2022Achterberg et al. 2018). Moreover, tree-ring growth and isotopic patterns measured from subfossil pinewood can reveal the conditions under which the trees once grew and provide dendrochronologists with information on abrupt changes in climate (Grudd et al. 2000;Leuschner et al. 2002;Moir 2012;Edvardsson et al. 2014a, b;Torbenson et al. 2015).
In Fennoscandia and adjacent areas, several tree-ring studies illustrating depositional histories of peatland trees have been published; however, these analyses focus predominantly on sites from southern Sweden (Edvardsson et al. 2012a(Edvardsson et al. , b, 2014a and the Baltic region (Pukien_ e 1997(Pukien_ e , 2001Edvardsson et al. 2016a). Much less attention has been paid to retrieving subfossil remains of peatland pines from high-latitude sites in northern Fennoscandia. In fact, in this region, dendrochronological research has focused particularly on subfossil pine remains unearthed from lake sites, which have produced large collections of tree-ring dated subfossil pinewood samples (Eronen et al. 1999a(Eronen et al. , b, 2002Grudd et al. 2002; bordering the lakeshore, after which the downed boles have been preserved in anoxic sedimentary environment of small Arctic lakes (Eronen et al. 1999a;Helama et al. 2010b). While the environmental conditions leading to tree accumulation in peatland and such lake sites may be much the same, their relative roles seem to vary between the settings; in detailed view the two types of sites may exhibit contrasting patterns of pinewood accumulation, as a result of diverging ecological conditions and taphonomical processes altering the preservation potential (Helama et al. 2017). Moreover, the climate of southern Fennoscandia and Baltic countries is much milder in comparison to that of the Arctic Circle. Consequently, the factors behind the ecological and sedimentological processes shaping the sedimentary record of peatland pines may vary considerably. Combined, there appears to be a need for a more northern focus in research efforts.
In this study, we concentrate on an assemblage of subfossil pine remains from a peatland site in the region of Enonteki€ o in Finnish Lapland. The high altitude and latitude pinewood findingswere dated using 14 C and treering methods to determine the phases when the species was present in this site, located today several tens of kilometres north of the coniferous timberline. To do so, we aim to disentangle the factors behind the establishment and demise of the population. Multiproxy comparisons are carried out between the palaeobotanical records of pine chronologies and palaeoclimatic reconstructions of temperature and hydroclimatic variations available from the region and adjacent areas.

Site of excavation
Pousuj€ arven suo (latitude 68°51 0 N, longitude 21°10 0 E) is a peatland site located in the Enonteki€ o region in northwest Finnish Lapland, at an altitude of 450 metres above the present-day sea level (Fig. 1). The region belongs to the sub-alpine birch forest zone (Ahti et al. 1968;Kauhanen 2013). Arboreal vegetation is characterized by mountain birch (Betula pubescens ssp. czerepanovii) forests and treeless tundra, between which dwarf birch (Betula nana) and bush-like willows (Salix spp.) occur invariable proportions. Mountain birch is the only forest-forming species around the site. Coniferous timberline lies~60 km south of our field site and is formed by Scots pine (Pinus sylvestris L.), where its presence can be explained by soil physics, the most important variable being soil water content (Sutinen & Middleton 2020). A 20-ton excavator, standing on the mineral soil adjacent to the peatland, was used to prepare a peat profile for the site. The information of the peat profile, adopted from the study of Sutinen et al. (2007), where an interpretation of the peat stratigraphy was presented in more detail, can be summarized as follows.
The profile extended from the surface of the peatland through the organic deposits. The top 0.6 m was composed of Carex-Sphagnum peat with gyttjahorizons containing dystrophic algae. Between 0.6 and 1.9 m, the peat was represented by an abundance of Equisetum as an indicator of increasingly wet conditions. This layer was divided into two sub-layers as follows. Between 0.6-1.3 m, the material was typified by Bryales-Sphagnum peat suggesting temporary drier periods. The sub-layer between 1.3 and 1.9 m was characterized by Carex-Sphagnum peat, indicating a stable fen-stage. The lowermost peat layer between 1.9 and 2.1 m was  Korhola et al. 2005), relative to the northern coniferous timberline formed by Pinus sylvestris (dotted line) (A) and to the present area of peatlands (both natural and drained) estimated for the 10 9 10 km 2 grid cells (Kaakinen et al. 2018) in northern Finland (B). composed of eutrophic Bryales-Sphagnum (H3) peat with an abundance of Menyanthes. The lower part of the profile, between 2.1 and 2.6 m, comprised gyttja deposits. The basal peat was dated to 7830AE80 a BP (8600 cal. a BP). The layer of gyttja rested on lacustrine silt. This excavation yielded 21 subfossil pine (P. sylvestris L.) trunks, with a concentration of findings at the depth of~1 m. The samples can be termed either megafossils (by virtue of their size) or subfossils (following the commonly accepted terminology for the Holocene/ Pleistocene remains of tree stems; see Helama et al. 2020 for discussion). Sample discs were cut in the field by a chainsaw. Eighteen well-preserved samples were analysed using 14 C and tree-ring methods. Three of the 14 C dates (Su-3595, Su-3596 and Su-3597) originate from the studyof Sutinen et al. (2007), the rest of the dates have not been published earlier.

C dating
Radiocarbon determination was carried out at three laboratories: the Radiocarbon Laboratory of the Geological Survey of Finland (M€ antynen et al. 1987;€ Aik€ a€ a et al. 1992), the Laboratory of Chronology, University of Helsinki ) and the Pozna n Radiocarbon Laboratory (e.g. Goslar et al. 2004). Among these, the laboratory at the Geological Survey of Finland no longer exists. First, the dates were given in conventional 14 C year BP (before AD 1950) with uncertainties pertaining to the counting statistics and activity measurements. Second, the conventional 14 C dates were calibrated into calendar years (cal. a BC) reported here with their two-sigma age range using the OXCAL (version 4.3) software (Bronk Ramsey 1995Ramsey , 2001. The Intcal13 calibration curve was used for calibration (Reimer et al. 2013). A cumulative probability density function (CPDF; Johnstone et al. 2006) was produced by OXCAL from the calibrated 14 C dates to illustrate the potential phases of increased accumulation of peatland pines in our site.

Tree-ring analyses
Cross-sections of the subfossil pine samples were cleaned, sanded and scanned using a flatbed scanner. Tree-ring widths were measured from the digital images using the CooRecorder 7.7 software (Rydval et al. 2014;Wilson et al. 2014;Maxwell & Larsson 2021). Trials to cross-date the resulting time series of ring widths against each other and against the existing master chronology built previously for northernmost Finnish Lapland (Eronen et al. 2002;Helama et al. 2008Helama et al. , 2019 were conducted. Cross-dating was carried out both visually and statistically (Holmes 1983). Tree-ring widths were transformed into dimensionless indices by fitting a 32-year spline function (Cook & Peters 1981;Cook et al. 1990b) to the series and dividing the observed values of ring widths by the values of the curve. Moreover, the series were pre-whitened to remove the remaining autocorrelation (Cook et al. 1990a). Pearson correlations were calculated between the sample series and between the sample series and the master chronologies. Moreover, the t-value (Baillie & Pilcher 1973) was used to quantify the agreement between the sample series and the master chronology. Generally, values of t > 3.5 may be taken as indicative of an acceptable dating (Baillie & Pilcher 1973). This threshold was also accepted in cross-dating tree-ring series of peatland pines (P. sylvestris) from northern Scotland (Moir et al. 2010;Moir 2012). The calculation of the t-value was carried out using the spline-detrended and pre-whitened index series.

C wiggle-matching
Multiple radiocarbon dates were used for 14 C wigglematching from subfossil pine samples and each provided one 14 C date. A set of tree-ring series cross-dated against each other provided a built-in time scale with known age differences between the radiocarbon dates. This made it possible to match the patterning of the 14 C dates to the 'wiggles' of the calibration curve improving the dating precision and accuracy Galimberti et al. 2004). An agreement index A was used as a test for model consistency, with an acceptance threshold of 60% (Bronk Ramsey 1995). For this purpose, the IntCal13 calibration curve was applied (Reimer et al. 2013) using OXCAL software (Bronk Ramsey 1995.

Reporting the geochronological ages
This study contains age determinations based on 14 C dating and tree-ring methods. The calibrated 14 C ages are reported as 'cal. a BC' throughout the paper. Since the tree-ring dates are not calibrated, they are reported simply as 'a BC'.

Postglacial land uplift
The study region occurs within the area of glacioisostatic rebound of the Earth's crust (Ekman 1997) and, as a consequence, the trees did not actually grow as far above the sea level as the altitudes of their subfossil remains indicate today. Subsequent to dating, the present-day altitude was adjusted for land uplift. In this study, a technique identical to that of Dahl and Nesje (1996) was used based on the sea level curves from Lygnen (Corner & Haugane 1993) and Pello-Rovaniemi (Saarnisto 1981).

Temperature estimates
The climatic conditions at the Pousuj€ arven suo site were adopted from a spatial model (Ojansuu & Henttonen 1983). This model was used to estimate the long-term BOREAS (1961-1990) mean July temperature (°C) and temperature sum (degree-days), as interpolated for each site based on their coordinates, continentality and altitude. These variables were selected as they have been previously found to be related to the germination of pine seeds (Henttonen et al. 1986) as well as the occurrence (Norokorpi 1994;Mikkola & Virtanen 2006) and growth (Helama et al. 2010a(Helama et al. , 2012 of northern pines.

Multiproxy data set
Peatland pine records from the Pousuj€ arven suo site were compared with several types of proxy data from the Enonteki€ o region. Such data from the Enonteki€ o region originate from the depositional history of 14 C and treering dated pine megafossils of lake sites in the region (Eronen 1979;Eronen et al. 1999a, b;Helama et al. 2004Helama et al. , 2010b. As previously suggested, the indications of past tree line and timberline variations were drawn from the altitudinally highest and second highest megafossil finds (Helama et al. 2010b), these records being adopted from the previous palaeoecological comparisons presented for the region (Helama 2016). Moreover, the availability of megafossils was taken to reflect the temporal variations in the density of pine population in the region (Helama et al. 2005). Comparisons were also made with pollenbased July temperature and annual precipitation reconstructions (Sepp€ a & Birks 2002) and pollen influx data (pine) from the Lake Toskaljavri (Sepp€ a et al. 2002) site (see Fig. 1 for site location). A threshold value of 500 grains cm À2 a À1 for the presence/absence of pine in the immediate vicinity of the site (Hyv€ arinen 1975; Hicks & Hyv€ arinen 1999; was accepted. In addition, the tree-ring (Helama et al. 2010a) and pollenbased July temperature reconstructions, and their amalgamated record representing sub-Arctic sites from the northernmost regions of Norway, Sweden, Finland, and Kola Peninsula (Helama et al. 2012), were compared with the peatland pine records from the Pousuj€ arven suo site. Hydroclimatic reconstructions inferred from cladoceran evidence, demonstrating the past fluctuations in lake levels in the Jierstivaara, Isohattu and Kuttanen sites (Korhola et al. 2005; see Fig. 1 for site locations), were also included in this multiproxy comparison. All these records were available at least over the Middle and Late Holocene including the past 7500 years.

C dating
Radiocarbon dating of subfossil pine trunks suggested that the samples probably represent two separate phases of subfossil tree deposition. That is, three of the samples showed calibrated 14 C ages older than 4500 cal. a BC, whereas the dates of the remaining pine samples were predominantly younger than 4000 cal. a BC (Table 1).
The cumulative probability density function (CPDF) of 14 C dates reinforced this view illustrating two accumulation pulses over~4800-4500 and 4000-3500 cal. a BC (Fig. 2), respectively, referred to hereafter as the peatland pine phases I and II. We also note that the shapes of the curves illustrating the pine phases were fairly similar, showing a rapid increase (peaking at 4700 and 3900-3800 cal. a BC) followed by a more gradual decline in the distribution. Clearly, these estimates together suggest a gap of four to five centuries between the two pine phases and an age difference of c. 800 years between their peaks.

Tree-ring analyses
Guided by 14 C dating results, the sample series were comparedwith each other and against the existing master chronology. As the samples appeared to represent two separate phases (Fig. 2), the samples within each phase were comparedwith each other. Moreover, the calibrated age ranges provided means to truncate the existing master chronology separately for each sample (Eronen et al. 2002;Helama et al. 2008Helama et al. , 2019; here, the sample series from the earlier and later phases could be reasonably compared with the data of the existing chronology truncated to represent the 4900-4400 and 4100-3400 cal. a BC intervals, respectively. In so doing, the tree-ring series representing the samples POU010 (t = 4.48) and POU011 (t = 7.24) showed notably high tvalues and their dating positions suggested an 88-year overlap (3967-3880 a BC) over which period the two series were correlated with r = 0.676 (t = 8.46) (Fig. 3). Using the mean record of these two series as a preliminary master chronology of our peatland site resulted in three additional potential dating positions for tree-ring series from the more recent pine phase. That is, the sample series POU003, POU004 and POUxxx showed t values in excess of 3.5, when correlating them with the mean record of all other remaining time series (Table 2; for visual comparisons, see Fig. 4). As a result, this chronology of five pines covered the 3986-3873 a BC period. Over this period, the new mean chronology correlated with r = 0.348 and t = 3.92 as compared with the existing master chronology (for visual comparison, see the lowermost plot in Fig. 4). In addition, there were four sample series (POU001, POU006, POU016 and POU020) showing t-values higher than 3.5 when dated against the existing master chronology. Accordingly, altogether nine samples could be dated by means of treering dating (Table 2). 14 C wiggle-matching As the age differences between the subsets of radiocarbon dated tree-ring sequences became precisely known, these 14 C data could be matched to the shape of the Intcal13 calibration curve (Reimer et al. 2013). The solution of the wiggle-match procedure demonstrated a patterning of the nine 14 C dates superposed on the curve (Fig. 5). Moreover, the wiggle-matching approach narrowed the uncertainty limits of the 14 C dates considerably (Table 2). Overall, the agreement between the samples and the curves, based on the A-index of Bronk Ramsey (1995), was 71.4%. The tree-ring dates fell within the two-sigma age ranges obtained from wigglematching.
Presence of pine 14 C and tree-ring dates of pine remains from the Pousuj€ arven suo site were consistent with the previous subfossil evidence from the Enonteki€ o region demonstrating high-altitude presence of pine coeval with our findings (Fig. 6A). The altitude of the previous subfossil pine findings from lake sites (560 m above present-day sea level) was higher than that of the Pousuj€ arven suo site (450 m above present-day sea level). Comparison with the accumulation curve of the subfossil pine remains from lake sites showed that the peatland pine phase I predated the surplus of lake site subfossil in the same region (Fig. 6B). The crest of the CPDF was coeval with a transient minimum in the lake site curve, around 4700 cal. a BC. It also seems that the peatland pine phase II started when the accumulation of lake site subfossils declined, and that the maximum of the phase II coincided with a local minimum of the lake site accumulation curve~3800 cal. a BC. Again, the end of Table 1. Pinus sylvestris samples with their code names, laboratory codes from 14 C dating (Lab. code), the depth (D) of the sample position (ND = not determined in the field), the number of tree rings measured from the sample discs (N), the number (n) and position (pos) of rings used for 14 C dating (number of n inner (inn) or outer (out) rings or ten middle (mid) rings counted onwards from the nth ring), the 14 C age determination in years BP ( 14 C age (a BP)), and the calibrated two-sigma age range in cal. a BC years (2-sigma (cal. a BC)).

Sample
Lab . (Eronen et al. 2002;Helama et al. 2008Helama et al. , 2019 and the comparison of the POU010 (grey line) and POU011 series (black line). Agreements between the time series are quantified using Pearson correlation (r) and t values of Baillie and Pilcher (1973).
the peatland pine phase II appears to coincide with the highest values in the lake site subfossils. Similar patterns in the accumulation curve of subfossil pines from lake sites in Swedish Lapland (Grudd et al. 2002) suggest that the comparison (not shown) could be extended over those more western sites. Compared to pine pollen influx from the lake site (704 m above present-day sea level), the peatland pine phases appeared to post-date the highest influx values that occurred at that site between circa 8500 and 5000 cal. a BC (Fig. 6C). The influx values indicated that pine could have been present at that site between 8500 and 2500 cal. a BC, during which period the influx values stayed around 500 grains cm À2 a À1 , this interval nevertheless overlapping with the peatland pine phases I and II. Interestingly, the influx curve exhibited a transient drop, around 3000 cal. a BC, following the peatland pine phase II.

Climatic fluctuations
Comparisons with palaeoclimatic evidence from sites adjacent to Pousuj€ arven suo provided a long-term context for the peatland pine evidence. Generally, the July temperature reconstruction based on pollen data demonstrated temperatures higher during the peatland Table 2. Dating results for the Pinus sylvestris tree-ring sequences achieved by means of dendrochronological cross-dating and 14 C wigglematching. The samples are given with their code names, Pearson correlation between the tree-ring sample series and the averaged chronology of remaining series (r master ), t-value of Baillie and Pilcher (1973), the cross-dated positions for the first (F tree ) and last (L tree ) measured rings in the sample, the corresponding rings in the 14 C sample used for wiggle-matching (F WM , L WM ), the two-sigma age range from wiggle-matching ( 14 C WM) and the agreement index A of Bronk Ramsey (1995) to quantify the wiggle-matching. Underlined values of r master and t represent cross-dating against the existing master chronology (Eronen et al. 2002;Helama et al. 2008Helama et al. , 2019, the other values of r master and t represent cross-dating against the local master chronology built in this study.   (Table 2) and the heights the error ranges from 14 C dating (Table 1). Black and grey curves show the Intcal13 14 C age (a BP) and its error, respectively. pine phases I and II than those obtained for the region today (Fig. 6D). Considering temperature fluctuations of shorter term, the timing of the two pine phases determined for the Pousuj€ arven suo site appear to overlap with transient intervals of ameliorated thermal conditions, inferred from pollen and tree-ring evidence.
These fluctuations represent changes towards warmer July temperatures. These intervals show high temperatures peaking at~4700 and 3800 cal. a BC. Comparisons with hydroclimatic fluctuations, inferred from cladoceran evidence, showed that the peatland pine phases I and II occurred during the period of low lake levels, as reconstructed for the study region ( Fig. 7A-C), as previously determined between 5000 and 2000 cal. a BC (Korhola et al. 2005). The peatland pine phase I appeared to coincide with the timing of lowest lake levels in the Jierstivaara site (Fig. 7A), whereas the peatland pine phase II was coeval with the lowest lake levels as recorded for the Kuttanen site (Fig. 7C). Moreover, the event of low precipitation in the pollenbased reconstruction (Fig. 7D) was coeval with the peatland pine phase I and that of high precipitation postdated the peatland pine phase II (Fig. 7D).

Estimating the Middle Holocene warmth
The effect of postglacial land uplift on the peatland pine phases I (4800-4500 cal. a BC) and II (4000-3500 cal. a BC) was estimated to be~30 and 25 m, respectively. Using a temperature lapse rate of 0.49°C per 100 m (Laaksonen 1976) this change (25-30 m) translates into a mid-summer temperature difference of 0.12-0.15°C between the Middle Holocene and the present day, that is, the mean temperature of the sites would not be more than a tenth of a centigrade warmer due to land uplift alone, in comparison to the present thermal conditions. Calculated from annually resolved temperature estimates, originating from a fusion of pollen and tree-ring data (Helama et al. 2012), mean July temperatures 2.0 and 1.7°C warmer than the present-day (AD 1961(AD -1990 conditions characterized the peatland pine phases I and II, respectively. These estimates suggest a midsummer temperature of around 13°C or higher when pine was present at the Pousuj€ arven suo site, this value markedly exceeding the threshold of 12.2°C, of which isoline has been found to stand as a statistical model for the timberline of pine in the region (Mikkola & Virtanen 2006). Long-term (AD 1961-1990) growing season thermal conditions at the Pousuj€ arven suo site, adopted from a spatial model (Ojansuu & Henttonen 1983), were characterized by mean July temperature of 11.3°C and temperature sum of 502 degree-days (DD), with however large year-to-year variations with corresponding maximum (minimum) values of 14.5°C (8.5°C) and 712 DD (286 DD). In the region of Finnish Lapland, the timberline of pine can be generally expressed by means of a temperature sum of 600 DD (Norokorpi 1994). Alternatively, the timberline has been modelled by a temperature sum of 591 DD or a mean July temperature of 12.2°C (Mikkola & Virtanen 2006). In addition, the tree limit of pine has been expressed by 550 DD (Norokorpi 1994). These estimates markedly exceeded BOREAS those obtained today in recent times (AD 1961(AD -1990 for the Pousuj€ arven suo site.

Discussion
Depositional history of the studied Pinus sylvestris trees illustrated two distinct phases of pine establishment dated to Middle Holocene intervals 4900-4400 and 4100-3400 cal. a BC. The subfossil specimens analysed represent downed boles (i.e. megafossils), which means their local origin is certain, as distinguished from microfossil or macrofossil findings that may have been transported prior to final deposition (Edvardsson et al. 2022). That is, the potentially transported materials may not be specific to the actual ecosystem, which in the case of our samples represents peatland. According to Edvardsson et al. (2014b), peatland wood specimens may in any case also originate from trees from drier marginal ground of the peatland basin as such trees may have either fallen into the adjacent peatland or become embedded in peat during peatland expansion. Primarily, the 14 C and tree-ring dated materials from the Pousuj€ arven suo site contribute to the timberline history of northern pines. The present-day site characteristics (Ojansuu & Henttonen 1983) demonstrate~1°C and 100 DD colder conditions than those observed commonly at the northern timberline of pine in the same region (Norokorpi 1994;Mikkola & Virtanen 2006). In any case, both temperature and hydroclimatic factors appear to have played roles of varying importance. First, the temperature difference (inferred from pollen data) between the Middle Holocene and present-day conditions ( Fig. 6) appeared to exceed the difference between the present-day temperatures at the timberline and at our study site. Pine recruitment depends on the seed crop, which, in turn, is related to summer warmth, especially in sub-Arctic sites. According to Henttonen et al. (1986), a temperature sum of 890 DD is required for 50% of the pine seed crop to be germinated. Palaeobotanical evidence demonstrating higher than present tree line and timberline positions during the Middle Holocene (Fig. 6) shows the potential for high-altitude pine recruitment that, in addition to seed germination, is limited also by other climatic factors such as snow cover and wind conditions that may affect especially the seedlings (Kullman 1981(Kullman , 1995 and maybe modified ground vegetation dynamics . Combined, these lines of evidence indicate that the studied pines were surviving at the Pousuj€ arven suo site in conditions below the contemporary tree line of the species. Yet, the conditions have since then become unsuitable for pines to recruit, and the site is located 60 km (see Fig. 1) north of the coniferous timberline formed today by the same species (Juntunen et al. 2002;Helama et al. 2020). Generally, these indications appear to link the occurrence of the peatland pines at the Pousuj€ arven suo site, and the subsequent southward retreat of the pine tree line, to the orbital forcing due to Milankovitch cycles (Berger 1978(Berger , 1988. In this respect, the findings concur with previous palaeobotanical works demonstrating an association between the high-latitude summer-temperature cooling due to the decreasing insolation and the retreat of pine and other arboreal species from high altitude and latitude sites in Fennoscandia through the Middle and Late Holocene  (Korhola et al. 2005), pollen-based reconstruction of annual precipitation from Lake Toskaljavri (Sepp€ a & Birks 2002) (D), and the cumulative probability density function (CPDF) for the 14 C dates of the peatland pine megafossils (this study) (E). The vertical dashed lines enclose the period of low lake levels based on the cladoceran records (Korhola et al. 2005). See Fig. 1 for site locations. (Kullman 1992(Kullman , 1995Selsing 1996;Eronen et al. 1999b;Sepp€ a & Birks 2002;Helama et al. 2004Helama et al. , 2022. Such temperature-mediated association may not be similarly evident at lower latitudes. Compared to peatland pine findings from more southern sites, however, we were not able to find extensive horizons of in-situ tree remains such as those described by Leuschner et al. (2007). That is, the phases of germination and mortality could not be reasonably inferred owing to low replication of our pine record. In common with studies from lower latitudes, in any case, the results from the Pousuj€ arven suo site appear to indicate fluctuations between drier and wetter periods.
In detailed view, the crests in the peatland pine CPDF curve appear to coincide with temporary lows in the supply of pinewood from lake sites, around 4700 and 3800 BC. While we must consider the dating uncertainties inherent to 14 C dating, the situation also mimics that observed previously for the depositional histories of Late Holocene pine trees on peatland and lake sites in southern Finland. In that setting, the accumulation of peatland pines was found to be high, but that of riparian trees low around AD 1000 whereas the accumulation of peatland pines declined while that of riparian trees increased around AD 1300 (Helama et al. 2017). It appears that the divergent courses of depositional histories obtained for the abovementioned southern sites can be explained by coinciding changes in the reconstructed soil moisture (Cook et al. 2015) and in the lake levels (Luoto 2009;Nevalainen et al. 2011;Nevalainen & Luoto 2012). While the dry conditions around AD 1000 coincided with a surplus of peatland pine availability, increasing moisture levels likely did not allow pine regeneration thereafter, due to excessively wet surfaces. Instead, the rising lake water level submerged the trees, recruited near the lake shores during the period of low lake water level, as indicated by increasing tree accumulation at lake sites; further, the wet phase was indicated by decreasing accumulation of peatland trees (Helama et al. 2017). A similar interpretation can be applied to the peatland pine phases I and II that would, accordingly, represent phases of lowered moisture supply in the Enontekiö region (Fig. 7), as previously inferred from cladoceran records (Korhola & Rautio 2001;Korhola et al. 2005) and sediment properties (Virkanen 2000). Moreover, the peatland pine phase was followed by considerable increases in the pollen-based precipitation reconstruction values (Sepp€ a & Birks 2002). According to Edvardsson et al. (2016b), the preservation potential of peatland wood may be lower than that of trees that fell into lakes. This may partially explain the relatively small size of our subfossil collection.
Additional proof for a coinciding dry period in northern Fennoscandia was previously derived from sedimentary archives of another lacustrine sites. In NE Finnish Lapland, Lake Njargajavri dried out between around 6000 and 3000 cal. a BC, this phase being recorded as a hiatus in sedimentation (V€ aliranta et al. 2005;Sarmaja-Korjonen et al. 2006). In the same region, cladoceran datawere used to infer past variations in the depth of Lake Kipoj€ arvi. A decreasing trend is found during the period from 6000 to 2500 cal. a BC, this event representing a lake level change of 2 m (Siitonen et al. 2011). Decreasing lake levelswere also inferred from midge-based proxy data from Lake V arddoaij avri (also in NE Finnish Lapland) around 3500 cal. a BC , this event marginally overlapping with our peatland pine phase II. Based on pollen data and depletion of 18 O in lacustrine carbonates, attenuated precipitation amounts from the Early to Middle Holocene were interpreted to have allowed pine expansion into the Abisko area in northernmost Sweden (Sepp€ a & Hammarlund 2000). In addition to sedimentary records, tree-ring isotopes can provide precisely dated palaeoclimate reconstructions (McCarroll & Loader 2004). In Finnish Lapland, the first two millennia of a recently built tree-ring 13 C record were found to represent relatively clear skies with gradually increasing cloud cover from the start of the record at 5500 BC (Helama et al. 2018). All these findings support the interpretation of sedimentary evidence originally put forth by Hyvärinen & Alhonen (1994) suggesting a dry interval with lower lake levels in Finnish Lapland from 7000 to 2500 cal. a BC (originally published as 8000-4000 uncal. a BP). In agreement with our interpretation, Eronen et al. (1999a) found several subfossil pine trees in situ or their root system anchored deep in the shore bank even 2 m below the present lake level in the timberline region of northern Finland and juxtaposed these treering dated specimens with the phase of rising lake levels since the Hyv€ arinen-Alhonen drought event. That is, the rising lake levels increased the supply of subfossil pine trees while the recruitment of peatland pines was hampered by the wetting of their substrate. We are not aware that these fluctuations would have been linkedwith non-climatic factors such the effect of uplift on regional hydrology.
These findings indicate that the pine establishment in the Pousuj€ arven suo site was maintained by the described Middle Holocene drought. Moreover, the gap between the peatland pine phases I and II demonstrates that the drought, reconstructed previously from sedimentary records of coarser temporal resolution and lower dating accuracy, was not continuous but likely interrupted by an interval of several centuries with moister surface conditions between 4400 and 4100 cal. a BC. Likely, the two-phase climax of the event was reached when pine was able to establish on the peatland during the 4900-4400 and 4100-3400 cal. a BC intervals. The improved dating and characterization of the event also mean that the hydroclimatic 5.2 ka event, which possibly represents a transition in global climate regimes between 5.6 and 5.0 ka (Magny et al. 2006) and could have been at least preliminarily overlap with the more vaguely dated Hyv€ arinen-Alhonen drought event, now appears to largely post-date at least the climax of the drought as recorded by our data. Generally, these interpretations follow the supposition that past events of peatland pine recruitment have occurred during the phases of dry surface conditions (Moir 2008;Moir et al. 2010;Edvardsson 2013;Torbenson et al. 2015;Edvardsson et al. 2016b). In this respect, it is known that any decrease in soil moisture within the uppermost 30 cm of peat soil is directly encountered by the root system of trees growing on the site (Paavilainen 1966;Boggie 1972). Accordingly, the physical and chemical changes within this column of peat soil ought to play an essential role in determining the changing soil-plant interactions. This is the context in which the past moisture changes have likely affected the establishment and survival of Middle Holocene peatland pine trees in the Pousuj€ arven suo site, demonstrating the sensitivity of these proxy data with respect to providing evidence of past drought events. Due to dry substrate conditions, the role of fire cannot be completely ruled out in the establishment dynamics. It is known that not all the boreal forests in Scandinavia, even those growing on swamps, have functioned as true long-term fire-free refugia (H€ ornberg et al. 1995).
An interesting correlation can be made to subfossil pinewood collections from high-altitude sites in central Sweden. In those sites (~63°N), the sample depth of treering dated pine trees was found to increase in both peatland and lake sites around 4900-4800 and 3800-3600 a BC (Gunnarson 2008), both these periods showing at least partial overlap with our indications of increased supply of subfossil pines around 4900-4400 and 4100-3400 cal. a BC. In an extended regional view, connections between the inferred hydroclimatic events appear less evident. A major tree establishment phase dendrochronologically dated to 5200-4600 a BC, demonstrated by subfossil pine trees unearthed from a south Swedish (~56°N) peatland (Edvardsson et al. 2012b), at least partially pre-dates the peatland pine phases in our site, and an event in southern Sweden dated to 4300 cal. a BC, when the pine population seems to have more than doubled within a century (Edvardsson et al. 2012a), appears to have taken place during the gap in our peatland pine chronology. Moreover, a low water-table event on Scottish peatlands around 3200-3000 cal. a BC recorded by a subfossil tree-ring chronology (Moir et al. 2010), with intensive pine establishment between 3199 and 3130 a BC (Moir 2012), post-dates the peatland pine phase in our site. According to Moir (2008), pine had been present on similar high-latitude sites in Scotland (the Cairngorms and Rannoch Moor), based on the 14 C dates collected from the literature, roughly over the range 5200-3800 cal. a BC. Already the original authors (Hyvärinen & Alhonen 1994), however, delimited the Hyvärinen-Alhonen drought event spatially to the northern Swedish/Finnish sites, more humid conditions characterizing more southern areas (Digerfeldt 1988;Harrison et al. 1991). In any case, it cannot be excluded that our peatland pine phase II corresponds to the period of Baltic peatland tree establishment around 3800 cal. a BC (Edvardsson et al. 2016a). These examples do not dispute the longstanding view of time-transgressive hydroclimatic development between the north and south of the region (Hyvärinen & Alhonen 1994; Almquist-Jacobson 1995). It seems possible that some of the events of shorter duration may have been synchronous across the region. Further, the results emphasize the need for more detailed comparisons of subfossil tree populations from northern and southern peatland sites in Fennoscandia and the Baltic region, to analyse the data on temporal scales not available from sedimentary records of coarser resolution and lower dating accuracy. Such analyses would require new collections of peatland tree remains, especially from northern Fennoscandia, to be contrasted with findings from southern Fennoscandia, the Baltic region and Scotland, from where the amounts of unearthed materials have increased vastly recently (Pukien_ e 1997(Pukien_ e , 2001Moir 2008Moir , 2012Moir et al. 2010;Edvardsson et al. 2012aEdvardsson et al. , b, 2014aEdvardsson et al. , b, 2016aEdvardsson et al. , 2022.

Conclusions
Subfossil tree findings from treeless peatlands have long been important for Quaternary research, providing key evidence of long-term climatic changes. Much of the related research has so far concentrated on sites in southern Sweden, the Baltic region, Scotland, and central Europe. Focussing here on high altitude and latitude findings, subfossil Pinus sylvestris remains from NW Finnish Lapland demonstrate two distinct phases of peatland pine establishment dated to Middle Holocene intervals 4900-4400 and 4100-3400 cal. a BC, likely interrupted by an interval of several centuries with less favourable growth conditions between 4400 and 4100 cal. a BC. The site of excavation is located 60 km north of the coniferous timberline formed today by the same species. Thermal and hydroclimatic fluctuations were both found to have played roles of varying importance in the establishment of this pine population and its demise. Multiproxy evidence showed that the occurrence of the peatland pines at the Pousujärven suo site can be linked to orbital forcing (Milankovitch cycles). Moreover, the peatland pine phases were coeval with indications of lowered moisture supply in the same region, as previously inferred from other proxy records. This line of evidence is consistent with the interpretation that past events of peatland pine recruitment occurred during phases of dry surface conditions, which are particularly beneficial for pine colonization. In any case, the temperature-mediated association distinguishes the significance of high altitude and latitude subfossil tree findings such as ours from the peatland pine chronologies constructed for sites at lower latitudes. The findings emphasize the need to compare subfossil tree populations from northern and southern peatland sites over extended regions, for example, Fennoscandia, the Baltic region and Scotland, in advanced palaeoclimatic analyses.