Holocene wet shifts in NW European bogs: evidence for the roles of external forcing and internal feedback from a high‐resolution study of peat properties, plant macrofossils and testate amoebae

Two conspicuous wet shifts in the peat stratigraphy of Store Mosse in southern Sweden, associated with bog‐wide changes in vegetation and degree of peat decomposition, were analysed at high resolution. The bog‐surface wetness (BSW) proxy data (organic matter bulk density, C/N ratio, plant macrofossils and testate amoebae) highlight the importance of interactions between vegetation composition, microtopography and degree of peat decomposition, and show that the bog system operated consistently during the two wet shifts (dated to c. 2700 and 1000 cal a bp) despite different internal and external conditions. A sensitive bog‐system state, associated with a degraded microtopography and well‐decomposed surface peat with low hydrological conductivity developed during sustained dry conditions, probably contributed to the large BSW amplitudes registered. Comparable bog systems are expected to operate in the same way, and regionally high sensitivity that developed in response to atmospheric circulation changes may partly explain synchronous registration of wet shifts. The wet shifts in Store Mosse were attributed to solar and volcanic forcing, respectively, and wet shifts of similar magnitude registered in other NW European bogs are likely to also have been externally forced.


Introduction
In contrast to other types of peatlands that are hydrologically connected to their surroundings via surface and groundwater flow, ombrotrophic bogs are ideal as climate sensors because their hydrology is controlled only by the balance between precipitation and evapotranspiration, also known as effective precipitation (Charman, 2002). Palaeoclimatic interpretations are therefore usually based on peat-stratigraphic evidence of changes in bog-surface wetness (BSW), but the relative importance of precipitation and temperature for changes in BSW differs between regions. Precipitation is likely to be the primary driver of BSW changes in western Europe, and BSW records from this region are probably best interpreted in terms of the strength and latitudinal position of summer westerly airflow (Charman et al., 2009).
Holocene BSW records from NW European ombrotrophic bogs often show 'background' century-scale and lowamplitude variability punctuated by large-amplitude changes displaying a characteristic asymmetrical pattern with an abrupt shift towards wetter conditions (wet shift) preceded by a gradual change towards drier conditions, and regional data compilations indicate that some wet shifts are synchronously registered (Barber et al., 2004;Charman et al., 2006;Swindles et al., 2013). As early as the 1930s, Granlund (1932) proposed that the peat horizons characterized by an abrupt transition from well-decomposed to well-preserved Sphagnum peat, which he frequently observed in south Swedish bogs and referred to as recurrence surfaces, were evidence of abrupt and synchronous changes towards wetter conditions. This view was later challenged, but a re-analysis of peat-stratigraphic data suggests that some recurrence surfaces are synchronous within SW Sweden (Rundgren, 2008). Although regionally synchronous registration is a strong indication of external forcing, it is not clear to what extent internal bog processes contribute to the registration of externally forced wet shifts, or if wet shifts may result from internal bog processes alone.
A large-amplitude BSW increase registered throughout NW Europe around 2800 cal a BP constitutes the most wellestablished regionally synchronous wet shift (van Geel et al., 1996;Speranza et al., 2003;Barber et al., 2004;Charman et al., 2006;Mauquoy et al., 2008;Plunkett and Swindles, 2008;Swindles et al., 2013;Mellström et al., 2015;Słowiński et al., 2016). Moreover, this '2.8-ka event' corresponds to the Sub-boreal -Sub-atlantic transition, a prominent Holocene climatic shift within the classical Blytt-Sernander climatostratigraphic scheme (Blytt, 1881;Sernander, 1908), and one of the most conspicuous recurrence surfaces (Granlund, 1932). Because the wet shift around 2800 cal a BP coincides with a marked increase in atmospheric 14 C concentration, it is frequently linked to a decrease in solar activity (van Geel et al., 1996;Speranza et al., 2003;Mauquoy et al., 2008). Indeed, model simulations suggest that events of low solar activity induce changes in atmospheric circulation over Europe associated with southward displacement of North Atlantic storm tracks (Martin-Puertas et al., 2012), which supports that BSW increases during the '2.8-ka event' reflect responses to a regional precipitation increase.
Wet shifts may also be externally forced by large volcanic eruptions. Deposition of tephra particles are known to cause plant death through physical damage to leaves and inhibition of photosynthetic activity, and toxic elements leached from tephra reduce plant growth and increase mortality (Payne and Blackford, 2005;Payne and Egan, 2019). Differences in the sensitivity to these impacts between plant types and species could therefore induce changes in vegetation composition and thereby hydrological conditions. Moreover, increased accumulation of plant litter would reduce hydraulic conductivity and increase BSW. Tephra deposition itself is documented to induce higher BSW in a similar way (Crowley et al., 1994), but any tephra-related impacts are likely to be limited to areas with extensive tephra cover. Volcanic eruptions may, however, also influence BSW through acid precipitation and its effects on vegetation. Some eruptions produce large amounts of sulphur dioxide forming sulphuric acid aerosol that becomes deposited even at great distances from the source (Payne and Blackford, 2005; Payne and Egan, 2019; Thordarson and Self, 2003). Peatland vegetation changes have been attributed to acid deposition (Hogg et al., 1995;Kokfelt et al., 2016), and severe impacts on bog vegetation have been shown in field experiments (Payne and Blackford, 2005). Moreover, stratospheric input of sulphate aerosols from major volcanic eruptions is known to cause regional cooling that may last several years (Payne and Egan, 2019) and potentially influence BSW, but any impacts on precipitation patterns are likely to be more important.
Bog vegetation, microtopography, peat properties (degree of peat decomposition and hydraulic conductivity) and bogsurface hydrology are important components of a complex system characterized by numerous feedbacks (Charman, 2002;Belyea and Malmer, 2004;Morris et al., 2011Morris et al., , 2015Swindles et al., 2012). Because many of these are positive, peatland ecosystems may exhibit bistability and therefore go through regime shifts (Eppinga et al., 2009;Lamentowicz et al., 2019), which can help explain the abrupt character of wet shifts. Bistability occurs because Sphagnum mosses and vascular plants, the two dominant plant types in ombrotrophic bogs, influence their environment in ways that promote their own dominance (Malmer et al., 1994;van Breemen, 1999;Eppinga et al., 2009). For example, a BSW increase will directly promote Sphagnum growth, but vascular plants will also be replaced by Sphagnum because nutrients delivered by precipitation will to a larger extent be trapped by the expanding mosses and become less accessible to vascular plant roots. On the other hand, a BSW decrease will directly limit Sphagnum growth, while vascular plants remain to have access to water below the surface. In addition, decomposition of dry surface peat increases nutrient availability, and the expansion of vascular plants impedes Sphagnum growth through shading (Malmer et al., 1994;van Breemen, 1999;Eppinga et al., 2009). Feedback processes between peat properties and bog-surface hydrology are also likely to contribute to the sensitivity of bog systems (Morris et al., 2011(Morris et al., , 2015Swindles et al., 2012). For example, a prolonged period of dry conditions will result in a welldecomposed peat surface. If effective precipitation increases in this situation, the low hydraulic conductivity will promote a rapid increase in water level and BSW. The asymmetrical pattern observed in many BSW records may also reflect the way in which BSW becomes registered by the process of peat decomposition. During a sustained period of dry conditions, the water level will gradually move downwards, which results in decomposition of previously formed peat. As an effect of shorter exposure to aerobic conditions, the influence of this secondary decomposition (Borgmark and Schoning, 2005;Morris et al. 2015) becomes weaker with depth. Therefore, BSW will appear to have decreased more gradually than the underlying decrease in effective precipitation. More importantly, the signal of low BSW will pre-date the hydrological change experienced. Corresponding post-depositional alteration of the peat decomposition signal does not occur during an increase in effective precipitation. Moreover, the full amplitude of the hydrological change responsible for a wet shift may not become registered by peat decomposition proxies because excess water will be discharged as surface run-off.
In this study, we aim to provide a better understanding of the influence of external forcing and internal feedback on the registration of Holocene BSW changes in NW European ombrotrophic bogs, in particular wet shifts characterized by major changes in plant-community composition. To achieve this, we apply a set of biological, physical and chemical BSW proxy methods to the stratigraphy of Store Mosse in southern Sweden, known from previous studies to include two conspicuous wet shifts associated with major and bog-wide changes in vegetation and degree of peat decomposition (Svensson, 1988a(Svensson, , 1988b. Because of the unusually large size and uniform stratigraphy of this bog, these wet shifts are expected to reflect regional hydrological changes and thereby constitute suitable targets for the study of processes involved in the registration of NW European wet shifts. Primarily, we base our interpretations on high-resolution records of peat properties, plant macrofossils and testate amoebae obtained across the two wet shifts, focusing on similarities and differences in BSW registration by the different proxies, but we also place these records in the context of proxy data obtained from peat intervals outside the two wet shifts. Through radiocarbon dating of a large number of closely spaced samples, we set out to identify any synchronicity with well-established solar activity or volcanic events implying that the wet shifts were externally forced. With this approach we address the following questions: • Does the characteristic asymmetrical pattern of largeamplitude BSW changes (wet shifts) observed in Holocene bog records from NW Europe reflect the influence of external forcing or is it, partly or entirely, due to internal feedback processes in the bog system? • What are the roles of plant-community composition, bogsurface microtopography and degree of peat decomposition in any such feedback processes? • Which external forcing factors may have influenced BSW in NW Europe during the Holocene?

Site description
Store Mosse (57°15′N, 13°55′E) is the largest (almost 100 km 2 ) mire complex in southern Sweden, situated at 160-170 m a.s.l. in a low-relief area dominated by till, glaciofluvial and glaciolacustrine sediments overlying bedrock consisting mainly of Precambrian gneisses and granites (Persson, 2008) ( Figure 1). It is separated into three major ombrotrophic bog areas with marginal minerotrophic fens, and the southern and largest bog area is further split by a penetrating sand dune system and a network of fen areas. During the deglaciation around 14 000 cal a BP, the Store Mosse area was part of a major glacial lake characterized by sand deposition. Subsequent differential isostatic rebound resulted in lake drainage and dune formation (Persson, 2008). A lake continued to exist in the southernmost part of the southern bog area but transformed into a fen around 9000 cal a BP (Svensson, 1988b  transformed lakes into fens. A major road and a railroad cross the mire complex at the northern margin of the southern bog area (Figure 1), and large-scale peat mining occurred between 1905 and 1967 at Store Mosse, which has been a national park since 1982. Data collected by the Swedish Meteorological and Hydrological Institute show an annual mean temperature of 5.5°C and a total annual precipitation of 751 mm . Corresponding values for January are −3.5°C and 56 mm, and for July 14.9°C and 82 mm. Store Mosse is situated in a landscape dominated by forest, which forms a mosaic together with peatlands, lakes and agricultural land. Conifer forest with pine (Pinus sylvestris) and spruce (Picea abies) is most widespread, but birch (Betula pendula, B. pubescens), alder (Alnus glutinosa), aspen (Populus tremula) and oak (Quercus robur) are also important components of the surrounding forests. Only single stunted individuals of Pinus sylvestris occur on the major bog areas, which show typical hummock and hollow microtopography. Hummocks are dominated by Sphagnum rubellum, heather (Calluna vulgaris) and hare's-tail cottongrass (Eriophorum vaginatum), and Sphagnum fuscum, Sphagnum austinii and crowberry (Empetrum nigrum) are typically found on their higher parts (Svensson, 1988b). Hollows are characterized by Sphagnum medium (formerly known as S. magellanicum, which is the name used below to facilitate comparisons with previous studies at Store Mosse), Sphagnum papillosum, S. balticum, S. cuspidatum and white beak-sedge (Rhynchospora alba).
The peat stratigraphy of Store Mosse has been thoroughly studied regarding changes in plant communities and accumulation of peat, carbon and nitrogen (Svensson, 1986(Svensson, , 1988a(Svensson, , 1988bMalmer and Wallén, 1993, 1999Malmer et al., 1997Malmer et al., , 2011Belyea and Malmer, 2004). Based on plant macrofossil analysis of numerous cores from the southern bog area, Svensson (1988aSvensson ( , 1988b identified three stages in bog development following the fen-bog transition and named them after the dominant Sphagnum species recorded: the Fuscum bog stage, the Rubellum-Fuscum bog stage and the Magellanicum bog stage. The bog-stage transitions, marked by conspicuous well-decomposed peat layers, were consistently recorded at approximately 3 and 2 m depth and dated to c. 2400 and 1000-1200 cal a BP. Later studies have mainly concerned dust deposition and its relationship to changes in climate and peat accumulation (Kylander et al., , 2016(Kylander et al., , 2018 but also stratigraphic changes in peat decomposition and peat geochemistry (Hansson et al., 2013;Martínez Cortizas et al., 2021). Recently, Ryberg et al. (2022) explored external and internal factors driving vegetation changes at Store Mosse but focused on changes occurring on longer (multidecadal to millennial) timescales compared with the present study.

Core retrieval, correlation, documentation and sampling
Overlapping 1-m core segments were retrieved in May 2009 with a 7.5-cm-diameter Russian sampler from the central part of the southern bog area of Store Mosse (57°14′8.60N, 13°55′ 35.04E) (Figure 1), within a few hundred metres of cores E12 and E13 extensively investigated by Svensson (1988aSvensson ( , 1988b. Coring stopped in sandy silt at 5.10 m. Core segments were combined into a complete sequence (SM2009) in the laboratory through correlation based primarily on distinct stratigraphic boundaries at 4.49, 2.70 and 1.76 m. After photo documentation and description, sampling was made at 1-cm resolution (except for six samples spanning 1.5 cm) down to 3.45 m. Lower resolution (4-6 cm) sampling was made between 3.45 and 5.10 m, yielding in total 411 samples.

Analysis of peat properties
Organic matter (OM) content was estimated at 272 levels of the peat sequence through loss-on-ignition (LOI) by heating subsamples spanning the entire depth interval of each core slice first to 105°C and then to 550°C (Chambers et al., 2011). OM content was calculated from the weight loss during heating to 550°C. Organic matter bulk density (OMBD) is known to be higher for well-decomposed peat (Chambers et al., 2011;Loisel et al., 2014) and was therefore used as a proxy for the degree of peat decomposition. Weighing of fresh core slices, and estimation of their cross-section areas based on area measurements at representative levels and interpolation, allowed calculation of wet bulk density (BD), and dry BD was then calculated using water content data from the LOI analysis. OMBD was calculated for 280 samples by multiplying dry BD with OM content. C/N mass ratio was used as an additional peat decomposition proxy based on the observation that decomposition primarily involves C loss, resulting in lower C/N mass ratios for well-decomposed peat (Kuhry and Vitt, 1996;Malmer et al., 1997;Loisel et al., 2014). Mass ratios were derived from C and N percentage data obtained for 309 subsamples through combustion in a Costech ECS 4010 elemental analyser.

Plant macrofossil analysis
The main peat components were determined for 141 samples by screening 1-2 cm 3 of material mixed with deionized water in a Petri dish under a stereo microscope. Macrofossil identification was aided by comparison with Nilsson (1952), Mauquoy and van Geel (2007) and modern reference material. Classification of components was inspired by the system applied within the Quadrat Leaf Count technique (Barber et al., 1994), and the following taxa were identified: Sphagnum section Acutifolia, S. section Cuspidata, S. magellanicum, S. spp., and Monocotyledons (Monocots). In addition, the amount of unidentifiable organic matter (UOM), i.e. plant material decomposed beyond recognition, was estimated and used as a supplementary peat decomposition proxy. The encountered S. section Acutifolia remains are assumed to mainly belong to S. fuscum and S. rubellum, as suggested by the dominance of these species in macrofossil records from adjacent coring sites at Store Mosse (Svensson 1986(Svensson , 1988a(Svensson , 1988b. Similarly, the finds of S. section Cuspidata are assumed to mainly represent S. cuspidatum. E. vaginatum remains dominate the Monocot component of Sphagnum peat samples, while remains of other Cyperaceae dominate in fen peat samples. Abundance of the main peat components was estimated on a scale from 1 (rare) to 5 (abundant) with respect to their contribution to the volume of each sample. The presence of sand particles and wood fragments was also noted. The single wood fragments encountered are believed to mainly represent twigs, but possibly also roots, of dwarf shrubs. BSW changes were inferred from the obtained records of UOM, Monocots and Sphagnum taxa characteristic of different bog microhabitats.

Analysis of testate amoebae
Testate amoebae were analysed in 29 samples taken immediately below and above the two distinct stratigraphic boundaries at 1.76 and 2.70 m to detect rapid changes in BSW. Testate amoebae were extracted from the peat as described by Charman et al. (2000) and analysed in microscope slides prepared from the 15-µm sieve residue. Identifi-cation follows the taxonomy of Charman et al. (2000) and Siemensma (2022). Interpretations were based on relative frequencies of taxa typical of low BSW (Bullinularia indica, Difflugia pulex and Trigonopyxis arcula type) and high BSW (Archerella flavum and Amphitrema wrightianum) (Charman et al., 2000;Amesbury et al., 2016).

Dating and construction of age-depth model
Material from 54 of the samples analysed for plant macrofossils was dated by accelerator mass spectrometry (AMS) at Lund University Radiocarbon Dating Laboratory. Samples were selected more closely around the distinct stratigraphic boundaries at c. 1.76 and 2.70 m, and within other intervals where peat stratigraphy and degree of decomposition suggested changes in accumulation rate. When possible, Sphagnum moss stems and leaves were picked out for dating. When only bulk peat could be used (17 samples), major rootlets and wood fragments were first removed (Table 1). To validate and complement the 14 C-based chronology, 15 samples from the interval 2.96-3.11 m, potentially containing some of the most important mid-Holocene isochrones in the Swedish tephrochronology (Wastegård and Boygle, 2012), were screened for tephra particles using the methods outlined by Pilcher et al. (1995). The peat was combusted at 550°C for 4 h and washed in 10% HCl. After mounting for microscopy in Canada Balsam, the concentration of tephra shards was estimated with the aid of added Lycopodium spores (Stockmarr, 1971) using tablets produced at the Department of Geology, Lund University (18 583 spores per tablet). Geochemical analysis providing concentrations of ten major oxides was performed using a Cameca SX-100 electron microprobe.
An age-depth model for the complete sequence was constructed using the P_Sequence option in OxCal v4.4.1 (Bronk Ramsey, 2008Ramsey, , 2009 and the INTCAL20 calibration dataset (Reimer et al., 2020) based on 53 radiocarbon ages (R_Date) and an input of 1999 ± 1 AD (−49 ± 1 BP) for 0.04-0.05 m (C_Date). This sample showed a 14 C activity above the 1950 level (Table 1) characteristic of the subsequent 'bomb pulse' period, and the alternative age of 1954 AD (−4 BP) obtained through calibration with CALIBomb (Reimer et al., 2004) and the Levin dataset (Levin and Kromer, 2004;Levin et al., 2008) was disregarded because immediately overlying living Sphagnum suggests a very recent age. The model was further constrained by boundaries at 0.00 and 5.10 m, and the Hekla-S/Kebister tephra identified at 3.035 m (C_Date) (see below). Because the identification of the Hekla-4 tephra at 3.095 m (see below) did not provide any age information additional to that available from radiocarbon dating, it was not used as an input for the age-depth model. The parameter k (number of accumulation events per unit depth) for the P_Sequence was set to 10 (10 events per metre). Linear interpolation between modelled mean (µ) ages for the dated levels was used to assign an age to each sample in the sequence, and minimum and maximum 2σ sample ages were assigned in the same way. Ages for boundaries between samples were also assigned through linear interpolation and used to calculate sample deposition times that were divided by sample increment to provide estimates of peat-growth rate (for convenience expressed in mm a −1 rather than m a −1 ).

Results and interpretation
Complete sequence Chronology, peat stratigraphy and peat-growth rate Two tephras derived from eruptions of the Hekla volcano in Iceland were identified in the screened samples ( Table 2).
The Hekla-4 tephra (4260 ± 20 cal a BP; Pilcher et al., 1995) was found at 3.095 m depth (c. 5700 shards g −1 wet peat), and the Hekla-S/Kebister tephra (3720 ± 30 cal a BP; Wastegård et al., 2008) at 3.035 m (c. 1640 shards g −1 wet peat). The reported age for Hekla-4 is consistent with the 14 C-based age of bulk peat from the same sample (4420-4149 cal a BP at 2σ) ( Table 1). Also the age reported for Hekla-S/Kebister is consistent with the 14 C data, which provide an age of 4422-4096 cal a BP for bulk peat sampled 2 cm below the tephra and 3555-3256 cal a BP for bulk peat sampled 6 cm above it (Table 1). An additional peak in tephra concentration was detected at 2.995 m (c. 655 shards g −1 wet peat). Only five analyses were successful, and three of these show a trachytic composition indicating an origin outside Iceland ( Table 2). The major element geochemistry is in very good agreement with proximal deposits of the P2 eruption of Sete Cidades in the Azores previously dated to c. 3800 cal a BP (Wastegård et al., 2020). This represents the most distantly recorded Holocene tephra from the Azores in NW Europe so far, but the low number of tephra shards analysed precludes correlation to a specific eruption. The identification of the two Hekla tephras strongly supports that a reliable chronology can be developed for the Store Mosse sequence through radiocarbon dating of carefully selected material, and the success of this approach is confirmed by the acceptable agreement index (A model = 61.8%, A overall = 66.8%) for the OxCal P_Sequence used to construct the age-depth model ( Figure 2).
Four peat-stratigraphic units separated by distinct boundaries were described for the 5.10-m sequence. A basal fen peat dominated by remains of vascular plants started to accumulate around 9900 cal a BP, directly on top of sandy silt assumed to have been deposited in the previous glacial lake ( Figure 3). This is superimposed by a unit of Sphagnum peat showing an overall upward increase in the degree of decomposition, and the transition from fen to bog at 4.49 m occurred around 5600 cal a BP. A second unit of Sphagnum peat, displaying a similar upward increase in the degree of decomposition, follows at 2.70 m (c. 2500 cal a BP). The uppermost Sphagnum peat unit, starting at 1.76 m (c. 900 cal a BP), shows a variable degree of peat decomposition above 9 cm and a 4-cm surface layer of living Sphagnum moss. The ages estimated for the three peatstratigraphic boundaries are broadly consistent with those determined for the transitions between the fen and bog stages in the southern bog area by Svensson (1988b), Kylander et al. (2013Kylander et al. ( , 2016Kylander et al. ( , 2018, Martínez Cortizas et al. (2021) and Ryberg et al. (2022). Very low peat-growth rates (generally <0.5 mm a −1 ) characterize the basal fen peat and the welldecomposed parts of the Sphagnum peat units, while the lower and more well-preserved parts of these units show high growth rates in the range 2.0-3.5 mm a −1 (Figure 3). The living mosses have grown at a rate of c. 4,5 mm a −1 .

Peat properties
OM content is uniform (98-100%) within the basal fen peat unit and overlying Sphagnum peat units, except for markedly lower values within the uppermost c. 0.2 m (Figure 3). OMBD is relatively high (0.09-0.12 g cm −3 ) in samples from the fen peat. Values within Sphagnum peat range between 0.03 and 0.15 g cm −3 , with generally higher values in the upper, visibly more decomposed parts of the lower two units. C content is relatively high (55-60%) in the upper part of the fen peat and varies between 43 and 57% in Sphagnum peat, with generally higher values in the upper parts of the lower two Sphagnum peat units. Also N content is relatively high (1.1-2.1%) in the fen peat. N values vary between 0.4 and 1.9% in Sphagnum peat and are generally higher in the upper parts of the lower two units. C/N mass ratio varies between 27 and 51 in the upper part of the fen peat, while the Sphagnum peat units show values within the range 25-127, with low values within their more decomposed parts and close to the surface. The obtained OMBD, C, N and C/N records are consistent with previous studies (Svensson, 1988b;Malmer et al., 1997;Kylander et al., 2013;Kylander et al., 2016;Martínez Cortizas et al., 2021;Ryberg et al., 2022). Table 1. Sample information, 14 C ages, individually calibrated age ranges and modelled age ranges for the 54 radiocarbon samples from the Store Mosse peat sequence. Major rootlets and wood fragments were removed from bulk peat samples before dating. Sample weight is the amount of C used for dating. Samples showing a 14 C activity above the the 1950 AD level are reported in the unit fM. Calibrated age ranges are maximum and minimum 2σ ages attained through individual calibration in OxCal v4.4.1 (Bronk Ramsey, 2008Ramsey, , 2009) using the INTCAL20 calibration dataset (Reimer et al., 2020), and CALIBomb (Reimer et al., 2004) and the Levin dataset (Levin and Kromer, 2004;Levin et al., 2008) for samples showing a 14 C activity above the 1950 AD level. Modelled age ranges are maximum and minimum 2σ ages derived from the age-depth model ( Figure 2) created using the P_Sequence option in OxCal v4.4.1 (Bronk Ramsey, 2008Ramsey, , 2009 and the INTCAL20 calibration dataset (Reimer et al., 2020). See text for further details about the age-depth model.

Plant macrofossils
The basal peat between 5.10 and 4.49 m (c. 9900-5600 cal a BP) is dominated by Monocots and UOM (Figure 3). Sand particles (not shown) and single remains of S. section Acutifolia were observed in the lowermost and uppermost samples, respectively. The pattern of changes in vegetation composition and degree of peat decomposition reflected by the plant macrofossil record (Figure 3) is remarkably similar to that documented by Svensson (1988aSvensson ( , 1988b for the same part of southern bog area, which allows correlation with the fen stage and subsequent three bog stages described by him. The absence of Sphagnum remains below 4.49 m identifies this as the fen-bog transition. The interval 4.49-2.74 m corresponds to the Fuscum bog stage, as indicated by the almost exclusive presence of S. section Acutifolia. Moreover, the high Monocot and UOM values between 3.01 and 2.74 m are consistent with a dominance of E. vaginatum remains and a high degree of peat decomposition in the upper part of the Fuscum bog stage (Svensson, 1988a(Svensson, , 1988b. The interval 2.74-1.76 m represents the Rubellum-Fuscum bog stage, as indicated by the continuous presence of S. section Acutifolia and the relatively high abundance of S. section Cuspidata in the lower part. In addition, the high Monocot and UOM values recorded between 2.30 and 1.76 m are consistent with the well-decomposed peat rich in E. vaginatum remains described as characteristic of the late Rubellum-Fuscum bog stage (Svensson, 1988a(Svensson, , 1988b. Consequently, the remaining part of the sequence corresponds to the Magellanicum bog stage, which is supported by the presence of S. magellanicum in the major part of this interval and the continuous presence of S. section Acutifolia. Moreover, the presence of single S. section Cuspidata remains in the lower part is consistent with the observation of thin Cuspidata or Cuspidatum peat layers within this bog stage  (Svensson, 1988a(Svensson, , 1988b. The plant macrofossil stratigraphy of SM2009 is also in general agreement with that presented by Ryberg et al. (2022) for a peat sequence retrieved approximately 1 km further south.

Testate amoebae
Testate amoeba assemblages between 2.86 and 2.77 m are characterized by Archerella flavum, Assulina muscorum, Figure 2. Age-depth model for the Store Mosse peat sequence, created using the P_Sequence option in OxCal v4.4.1 (Bronk Ramsey, 2008Ramsey, , 2009 and the INTCAL20 calibration dataset (Reimer et al., 2020), based on 53 radiocarbon ages and the age 3720 ± 30 cal a BP reported for the Hekla-S/ Kebister tephra (Wastegård et al., 2008) identified at 3.035 m. The model was further constrained by boundaries at 0.00 and 5.10 m and an input of 1999 ± 1 AD for 0.04-0.05 m, based on the elevated 'bomb pulse' 14 C activity recorded at this depth. Probability distributions for both individually calibrated (light grey) and modelled (dark grey) 14 C ages are shown. Dark and light grey bands are bounded by modelled 1σ and 2σ limits for the dated levels, respectively. See text for further details about the age-depth model. The inferred fen-bog transition and boundaries between the three bog stages described by Svensson (1988aSvensson ( , 1988b) are shown to the left. The boundaries defined for the two hydrological transition zones (LHTZ and UHTZ) are marked by dashed horizontal lines.

BSW changes across hydrological transition zones
Although the two bog-stage transitions described by Svensson (1988aSvensson ( , 1988b can be pinpointed to specific peat-stratigraphic levels (2.74 and 1.76 m), the records of peat properties and plant macrofossils show that the hydrological changes associated with these transitions occurred over centennial timescales, making them typical wet shifts (Figure 3). Here, we use the term 'hydrological transition zones' for the two peat intervals for which continuous data for peat properties (C/N and OMBD), and almost continuous plant macrofossil and testate amoeba data, are available with 1-cm resolution. Both hydrological transition zones were subdivided into three subzones (1-3) based on distinct patterns in the proxy data (Figures 4  and 5), and inferred BSW conditions are presented individually for each subzone below. To place the changes within the two hydrological transition zones (LHTZ and UHTZ) in a wider context, inferred BSW conditions are also presented for the peat intervals immediately below and above the two hydrological transition zones, although no testate amoeba data are available for these.

BSW changes across LHTZ
Pre-LHTZ (3.01-2.86 m, c. 3600-3100 cal a BP): low BSW (BSW minimum) All three peat decomposition proxies (C/N mass ratio, OMBD, UOM) indicate a very high degree of peat decomposition immediately below LHTZ ( Figure 5). This is supported by very  Figure 3. Peat-growth rate and records of peat properties and plant macrofossils (main peat components) for the complete Store Mosse peat sequence. Note that C/N mass ratio is shown on a reversed scale to facilitate comparisons with the other proxies of peat decomposition (organic matter bulk density and unidentified organic matter). See text for explanation of the semi-quantitative scale used for the main peat components. The inferred fen-bog transition and boundaries between the three bog stages described by Svensson (1988aSvensson ( , 1988b are indicated by solid horizontal lines. Dashed horizontal lines separate peat intervals for which plant macrofossil data are described individually in the text. The boundaries defined for the two hydrological transition zones (LHTZ and UHTZ) are marked by dotted horizontal lines.
high Monocot abundances. The total absence of identified Sphagnum remains is clearly an effect of poor preservation conditions and additional evidence of low BSW. Testate amoebae were not analysed.
Subzone 1 (2.86-2.77 m, c. 3100-2700 cal a BP): low BSW C/N, OMBD and UOM values indicate a degree of decomposition that is very high but possibly slightly lower compared with the peat interval immediately below ( Figure 5). Generally dry conditions and low BSW are supported by very high Monocot abundances and the Sphagnum record. Only S. section Acutifolia is identified in low abundances in the uppermost samples, but this taxon typical of relatively dry microhabitats is likely to have been present during the formation of the entire subzone, although this cannot be confirmed because of poor preservation conditions. Relatively high frequencies of the testate amoeba taxa B. indica, D. pulex and T. arcula type, relatively low frequencies of Archerella flavum, and absence of Amphitrema wrightianum strongly support generally low BSW.
Subzone 2 (2.77-2.70, c. 2700-2500 cal a BP): transition from low to high BSW The peat decomposition proxies indicate a moderate to high degree of decomposition, but also an overall decrease within the subzone ( Figure 5). OMBD values show a marked drop at the bottom of the subzone, UOM abundances reflect a rather gradual change to a moderate degree of decomposition throughout the entire subzone, and C/N values increase at the top. An overall BSW increase within Subzone 2 is consistent with a general decrease in Monocots and an accompanying increase in S. section Acutifolia, although abundances of this Sphagnum taxon are initially variable. The presence of S. section Cuspidata within the upper half of the subzone is the first clear evidence of peat formed by Sphagnum mosses characteristic of wet microhabitats. Testate amoeba assemblages are very different compared with the previous subzone and clearly reflect high BSW. B. indica is absent, and frequencies of D. pulex and T. arcula type are markedly lower. On the other hand, Archerella flavum frequencies are much higher, and Amphitrema wrightianum is present at low to moderate frequencies. Interestingly, this assemblage characteristic of wet microhabitats is not only found in the peat formed by a Sphagnum community characteristic of wet conditions but also in the welldecomposed peat found in the lower part of the subzone. Unexpectedly, this peat appears not to previously have been inhabited by testate amoeba taxa found in relatively dry microhabitats, as indicated both by the abrupt frequency changes of all testate amoeba taxa at the bottom of the subzone and the total absence of B. indica. This may, however, be explained by a situation with an extremely low water level, creating conditions that were too dry for the presence of testate amoebae in peat close to the surface. Moreover, a sustained period of extremely low water level may have resulted in secondary decomposition (Borgmark and Schoning, 2005;Morris et al., 2015), contributing to the very high degree of decomposition within the lower part of Subzone 2, and possibly also Subzone 1 and the peat interval immediately below LHTZ. This implies that the testate amoeba assemblage characteristic of wet microhabitats identified within Subzone 2 represents a community established as the water level subsequently moved upwards, and the abrupt assemblage shift at the bottom of the subzone may be an effect of the low hydraulic conductivity of the well-decomposed peat reached by the rising water level (Morris et al., 2011). However, low peat-growth rates may also contribute to the abruptly registered assemblage change, but these are likely to be influenced by secondary decomposition. Post-LHTZ (2.60-2.50 m, c. 2400-2200 cal a BP): high BSW All three decomposition proxies indicate a low degree of peat decomposition immediately above LHTZ ( Figure 5). Monocot abundances are, however, higher compared with Subzone 3, which suggests slightly lower BSW. This is supported by the Sphagnum data showing the discontinuous presence of S. section Cuspidata, continuous presence of S. section Acutifolia and low abundance of S. magellanicum in the uppermost sample. Testate amoebae were not analysed.

BSW changes across UHTZ
Pre-UHTZ (2.06-1.91 m, c. 1600-1200 cal a BP): low BSW C/N, OMBD and UOM values indicate a high to very high degree of peat decomposition immediately below UHTZ ( Figure 5). This is consistent with very high Monocot abundances. S. section Acutifolia is identified at low to moderate abundances. Although the Sphagnum record should be interpreted with caution because of poor preservation conditions, the presence of S. section Acutifolia supports the evidence from the other proxies of low BSW. Testate amoebae were not analysed. All three peat decomposition proxies indicate a very high degree of decomposition, and C/N and UOM values suggest that conditions were even slightly drier than during the formation of the peat interval immediately below ( Figure 5). Low BSW is supported by very high Monocot abundances. S. section Acutifolia is the only Sphagnum taxon identified, and its discontinuous presence and low abundances are likely  Figure 5. Selected records of peat properties, plant macrofossils (main peat components) and testate amoebae for the two hydrological transition zones (LHTZ and UHTZ) and adjacent intervals of the Store Mosse peat sequence. Note that C/N mass ratio is shown on a reversed scale to facilitate comparisons with the other proxies of peat decomposition (organic matter bulk density and unidentified organic matter). See text for explanation of the semi-quantitative scale used for the main peat components. The boundaries defined for LHTZ and UHTZ are indicated by solid horizontal lines. Dotted horizontal lines separate the three subzones of the two hydrological transition zones and delimit the peat intervals immediately below and above LHTZ and UHTZ, also described and discussed in detail in the text.
to primarily reflect poor preservation conditions and therefore support a slightly higher degree of decomposition compared with the underlying peat. The composition of testate amoeba assemblages, with relatively high frequencies of B. indica, D. pulex and T. arcula type, relatively low frequencies of Archerella flavum, and absence of Amphitrema wrightianum, is consistent with low BSW. Interestingly, the testate amoeba assemblage of the uppermost sample deviates by showing a markedly higher frequency of Archerella flavum. Because frequencies of the three taxa typical of low BSW remain relatively high in the same sample, the elevated frequency of Archerella flavum in the uppermost sample is best explained by a mix of assemblages characteristic of relatively dry and wet conditions. This probably reflects establishment of taxa preferring wet conditions in peat previously inhabited by taxa found in relatively dry microhabitats as the water level moved upwards in connection with the rapid water-level increase inferred for Subzone 2 (see below).
Subzone 2 (1.80-1.76 cm, c. 1000-900 cal a BP): transition from low to high BSW C/N, OMBD and UOM values indicate a change from very high to low degree of peat decomposition within the subzone, and this is consistent with a major decrease in Monocot abundances ( Figure 5). A transition from low to high BSW is supported by a marked increase in S. section Acutifolia abundances, which is likely primarily to be an effect of more favourable preservation conditions. The lack of evidence of S. section Cuspidata or S. magellanicum even in the uppermost part, where the degree of peat decomposition is low, suggests that a bog-surface vegetation characteristic of wet conditions was not developed. Testate amoeba assemblages indicate generally high BSW. B. indica and T. arcula type are absent, with exception of the lowermost sample analysed, and D. pulex occurs at markedly lower frequencies compared with Subzone 1. Moreover, Archerella flavum frequencies are very high, and Amphitrema wrightianum is present at low frequencies in the upper part. Similar to the uppermost sample of the previous subzone, the lowermost sample of Subzone 2 analysed clearly reflects a mix between testate amoeba assemblages characteristic of relatively dry and wet microhabitats. As inferred for LHTZ, an extremely low water level probably made it impossible for testate amoebae to live in peat close to the surface but, in contrast to LHTZ, taxa tolerating relatively dry conditions were able to survive in the lowermost peat belonging to Subzone 2. This was later replaced by an assemblage characteristic of wet microhabitats as the water level subsequently moved upwards, resulting in the observed assemblage mix. Although this mix is recorded over an interval of 3 cm at the top of Subzone 1 and bottom of Subzone 2, the associated BSW increase appears to have occurred very rapidly, probably as an effect of the low hydraulic conductivity of the well-decomposed peat (Morris et al., 2011). Low peatgrowth rates probably also contribute to the abruptly registered assemblage change, but these may partly reflect secondary decomposition. C/N, OMBD and UOM values indicate a low degree of peat decomposition ( Figure 5). Generally high BSW is supported by low Monocot abundances. Favourable preservation conditions suggest that the recorded abundances of Sphagnum taxa are representative of the vegetation composition. Although species belonging to S. section Acutifo-lia appears to have been dominant, the presence of S. magellanicum remains in some samples is the first clear evidence of peat formed in a wet microhabitat. Testate amoeba assemblages are similar to those found within the upper part of Subzone 2, but somewhat higher BSW is indicated by slightly lower frequencies of D. pulex and slightly higher frequencies of Amphitrema wrightianum.
Post-UHTZ (1.70-1.66 m, c. 900-800 cal a BP): high BSW A higher degree of decomposition compared with Subzone 3 is indicated by the C/N record, but this is not clear from the other peat decomposition records ( Figure 5). UOM values remain low, while the OMBD record shows only a minor increase. Slightly lower BSW is, however, supported by a relatively high Monocot abundance in the upper part of the peat interval. Although S. section Acutifolia appears to have been the dominant Sphagnum taxon, the identification of S. magellanicum and S. section Cuspidata in one sample shows that wet microhabitats existed. However, a slight post-UHTZ BSW increase is inferred from the C/N and Monocot data. Testate amoebae were not analysed.

Evaluation of BSW proxies
Reliability of the BSW signal The independent records of C/N mass ratio, OMBD and UOM show very good agreement, both for the complete sequence and for the two hydrological transition zones (Figure 3), confirming that these proxies reliably reflect the degree of peat decomposition and supporting their use as BSW indicators.
Our observation that most of the plant remains classified as Monocots belong to E. vaginatum, which is a species with tissues very resistant to decomposition (Svensson, 1986;Malmer et al., 1997) that mainly occurs in dry bog habitats at Store Mosse (Svensson, 1988b), supports the use of Monocot abundance as a complementary BSW proxy. However, some of the recorded E. vaginatum tissues are possibly rootlets penetrating down from individuals growing at a later stage. This could compromise BSW interpretations for some parts of the sequence, but rootlet penetration from above is unlikely to be a problem within LHTZ and UHTZ, because the progressively wetter conditions would have made it difficult for E. vaginatum to survive. Clearly, Sphagnum abundance estimates may be influenced by the lower preservation potential for Sphagnum remains during aerobic conditions, but this effect was considered when interpreting data showing an absence or low abundance of taxa typical of dry microhabitats (e.g. S. section Acutifolia).

Temporal resolution and setting depth of the BSW signal
When interpreting high-resolution BSW records, it is important to understand if the hydrological signals provided by the proxy methods applied are synchronous, or if minor lags occur because of different registration processes (Morris et al., 2015). Sphagnum remains record hydrological conditions experienced by the bog-surface community, and photosynthetically active (green) Sphagnum mosses are typically found within the uppermost c. 5 cm (Booth et al. 2010). Therefore, the BSW signal provided by Sphagnum remains represents a timeaveraged signal determined by moss growth rate. However, the bog-surface community response is also likely to occur with some lag associated with interaction between existing species and establishment of new species. Accordingly, Sphagnum abundance data generally reflect BSW changes occurring on multi-decadal timescales. The degree of peat decomposition is largely determined by the time that dead plant tissues are exposed to aerobic conditions, i.e. the time they remain in the acrotelm (Malmer et al., 1997;Chambers et al., 2011). Therefore, the degree of decomposition is controlled by the depth to the water table and the peat accumulation rate. However, because the water level of a bog fluctuates on seasonal to multi-decadal timescales, the main part of the decomposition takes place in the upper acrotelm only occasionally reached by the water table, i.e. typically within c. 5 cm of the bog surface. This means that the BSW signals in peat decomposition and Sphagnum data are likely to be essentially synchronous and have a comparable temporal resolution. Testate amoebae inhabit bog surfaces down to a depth of 10 cm or more (Mitchell and Gilbert, 2004). However, the main part of the testate amoeba population lives within a relatively narrow depth interval immediately below the surface layer of photosynthetically active Sphagnum mosses, and the assemblages occurring there are considered most representative of those found in the stratigraphic record (Charman et al., 2000;Booth et al., 2010). This implies that testate amoeba data can be used to detect BSW changes occurring on a shorter (decadal or even sub-decadal) timescale (Amesbury et al., 2012) compared with the signals of Sphagnum abundance and the degree of peat decomposition, but also that the BSW signal of testate amoebae is set at a slightly lower level, making it not fully synchronous with the signals of the other proxies.
Comparison of BSW changes across the two hydrological transition zones BSW registration by the different proxies is remarkably similar across LHTZ and UHTZ, which suggests that the Store Mosse bog system features a fundamental set of interacting processes with the capacity to operate consistently despite differences in bog community composition and climatic conditions. There are, however, minor differences in the BSW records ( Figure 5). A pre-LHTZ BSW minimum is suggested by the peat decomposition proxies, while a minimum during Subzone 1 is suggested for UHTZ. This is supported by the assumed preservation-dependent records of S. section Acutifolia, showing an absence of identified remains in the peat interval immediately below LHTZ and minimum abundances within Subzone 1 of UHTZ. Based on the absence of S. section Acutifolia remains, the lowest BSW experienced in connection with both hydrological transition zones occurred immediately below LHTZ. The evidence of increasing BSW within the transitional Subzone 2 differs in detail between LHTZ and UHTZ. Most importantly, the identification of S. section Cuspidata shows that peat was formed by Sphagnum mosses characteristic of wet microhabitats in the upper part of Subzone 2 within LHTZ, while the first evidence of similar habitats (S. magellanicum) within UHTZ is found at the bottom of Subzone 3. Moreover, the first testate amoeba assemblage indicative of high BSW is observed already below the Subzone 1/2 boundary within UHTZ. Considering all proxy records, maximum BSW conditions occur within Subzone 3 of both LHTZ and UHTZ. However, while moderate to high abundances of S. section Cuspidata in LHTZ clearly reflect the presence of a Sphagnum community characteristic of wet conditions, only low abundances of S. magellanicum are noted in some samples in UHTZ. This difference in the Sphagnum records may reflect relatively higher BSW within Subzone 3 of LHTZ. Slightly lower BSW is inferred for the peat intervals immediately above both hydrological transition zones but based on different evidence. For LHTZ, this is suggested by the Sphagnum and Monocot records but is not clear from the peat decomposition proxies. For UHTZ, lower BSW is suggested by the C/N and Monocot records, but still relatively wet conditions are reflected by low abundances of S. section Cuspidata and S. magellanicum in one sample. Although BSW was in part very low, the radiocarbon data are consistent with continuous accumulation of the peat belonging to LHTZ and UHTZ (Figure 2), suggesting that any secondary decomposition occurring during periods of extremely low water level was not strong enough to cause hiatuses, but this process may have resulted in lower (apparent) peat accumulation rates. Despite dating of densely spaced samples, the generally large uncertainties associated with radiocarbon dates make it difficult to estimate the time intervals represented by the two hydrological transition zones. Based on the age-depth model, the peat intervals represented by LHTZ and UHTZ span c. 700 and 300 years, respectively. Corresponding estimates for Subzone 2, where the major part of the BSW increase is registered, are c. 200 and c. 100 years.
Generally low BSW was also inferred for stratigraphic intervals corresponding to the onset of LHTZ and UHTZ by Ryberg et al. (2022) at their coring site c. 1 km further south based on peat decomposition and plant macrofossil data. Apart from frequent remains of E. vaginatum and other Cyperaceae, they found macroscopic charcoal fragments at some levels and concluded that the dry bog-surface conditions during these periods promoted recurring bog fires. This is consistent with plant macrofossil records from similar bog environments, showing that E. vaginatum is an early colonizer after severe bog fires (Tuittila et al. 2007;Sillasoo et al. 2011). However, no charcoal fragments were identified in the present study, suggesting that the fires detected by Ryberg et al. (2022) did not affect the entire southern bog area.

Evidence of internal feedback processes in the bog system
The peat decomposition proxies (C/N, OMBD, UOM and also Monocots) reflect a gradual BSW increase throughout Subzone 2, while the Sphagnum data reflect a BSW increase within the upper part of Subzone 2 of LHTZ and at the bottom of Subzone 3 within UHTZ ( Figure 5). Because the BSW signals of peat decomposition and Sphagnum abundance are set at a similar level (within the uppermost c. 5 cm of the bog surface) and have a similar resolution, the BSW registration by Sphagnum at a relatively higher level during LHTZ and UHTZ probably reflects the time it takes for taxa characteristic of wet microhabitats to establish or expand. Interestingly, the testate amoeba data reflect an abrupt BSW increase already at the bottom of Subzone 2. The fact that most testate amoebae live below the surface peat layer, where mosses are photosynthetically active and the main part of peat decomposition occurs, implies that the BSW signal of testate amoebae normally is set slightly later as decomposing plant tissues are gradually displaced downwards. However, as already pointed out, the observed presence of testate amoeba assemblages characteristic of wet microhabitats in well-decomposed peat within Subzone 2 (and the upper part of Subzone 1 of UHTZ) together with the limited evidence for a testate amoeba assemblage mix is incompatible with normal conditions of continuous peat accumulation and limited BSW variability. As suggested above, the data are best explained by a recovery of the bog water level following a sustained period of extremely dry conditions, because this water-level rise would be registered by testate amoebae slightly before conditions for peat decomposition and Sphagnum communities are changed at the bog surface. Moreover, the low hydraulic conductivity of the well-decomposed peat developed during the previously extremely dry conditions (Morris et al., 2011) would cause registration of an abrupt BSW increase by testate amoebae, as observed.
After a prolonged period of low water level, a bog surface is likely to be degraded with discontinuous vegetation cover. This is consistent with the data for the lower part of Subzone 2. Only species belonging to S. section Acutifolia appear to have been present, and the Monocot records indicate that E. vaginatum was an important component of the vegetation. Instead of a microtopography with hummocks and hollows, the bog surface was probably relatively smooth and characterized by a discontinuous Sphagnum cover, patches of decomposing peat, and scattered tussocks of E. vaginatum. Bog-wide surface degradation is confirmed by the conspicuous and welldecomposed peat layers rich in E. vaginatum remains observed at the stratigraphic positions of LHTZ and UHTZ throughout the southern bog area (Svensson, 1988a(Svensson, , 1988bRyberg et al., 2022). An increase in effective precipitation occurring in this situation will initially result in a water-level rise, but later any excess water would be efficiently discharged as surface run-off as an effect of the relatively smooth topography. Subsequently, hummock development in response to the BSW increase will, however, cause less efficient surface drainage and thereby contribute to higher BSW. At Store Mosse, hummock development was probably mediated by E. vaginatum tussocks offering structural support and suitable microclimatic and hydrological conditions for S. section Acutifolia (Tuittila et al., 2000;Pouliot et al., 2011). As indicated by the plant macrofossil data, continued differentiation of the bog surface into hummocks and hollows allowed establishment of Sphagnum taxa characteristic of wet microhabitats. Moreover, the low degree of decomposition of the peat within Subzone 3 is probably an effect of a well-developed microtopography capable of retaining water and maintaining maximum BSW conditions. A bog surface characterized by well-developed hummocks and hollows experiencing a sustained high level of effective precipitation is expected to maintain high BSW. Onset of a BSW decrease is, however, observed shortly above both LHTZ and UHTZ. Above LHTZ, the three peat decomposition proxies together reflect a trend towards lower BSW that lasted over a millennium and culminated within UHTZ, and generally dry conditions are supported by high Monocot abundances and total dominance of S. section Acutifolia (Figure 3). Remains of S. section Cuspidata and S. magellanicum are only present in one sample, indicating that the microtopography developed within LHTZ had degraded. Above UHTZ, on the other hand, there is no evidence of a long-term BSW trend. Instead, low-amplitude BSW fluctuations around a relatively constant mean value are indicated by the peat decomposition proxies and Monocot abundances ( Figure 3). Moreover, the rich occurrence of S. magellanicum remains indicates that a well-developed microtopography continued to exist for many centuries.
As suggested by previous studies (Belyea and Malmer, 2004;Morris et al., 2011Morris et al., , 2015Swindles et al., 2012) and the Store Mosse BSW records, vegetation composition, microtopography and the degree of decomposition of surface peat are key factors determining the sensitivity of an ombrotrophic bog system. Consequently, a specific change in effective precipitation is likely to be registered differently depending on when it occurs. For example, the large-amplitude BSW increases registered across LHTZ and UHTZ do not necessarily reflect exceptionally large increases in effective precipitation. They could also reflect disproportionally large responses of a bog system that was in a highly sensitive state resulting from a preceding dry period with a very low water level. Also the rate at which the increase in effective precipitation occurred may be overestimated in this situation, because BSW will increase more rapidly when the recovering water level reaches a welldecomposed surface peat with low hydraulic conductivity. As hydrological sensitivity may vary between ombrotrophic bogs within a region at any point in time, the registered amplitude of a specific wet shift is likely to be different. However, many bogs within a region are expected to simultaneously develop a similar degree of sensitivity as an effect of large-scale changes in atmospheric circulation. Accordingly, synchronous registration of a wet shift within a region does not necessarily imply that the associated increase in effective precipitation was particularly large. It may also reflect the widespread presence of highly sensitive bogs because of pre-conditioning during a preceding period of low effective precipitation. Importantly, an event of external forcing occurring in such a situation is more likely to become regionally registered as a major wet shift.
The age estimated for the bottom of Subzone 2 of UHTZ at 1.80 m is c. 1000 cal a BP (c. 1030-910 cal a BP). No exceptional changes in atmospheric 14 C concentration are reported at this time. However, the major fissure eruption Eldgjá in the Icelandic Katla volcanic system is dated to 939 AD/1011 cal a BP based on a major sulphur peak in Greenland ice (Sigl et al., 2015). The eruption is estimated to have emitted approximately 450 Mt of sulphuric acid aerosol over 3-8 years, making it the globally largest known basaltic flood lava eruption in the last millennium (Thordarson et al., 2001). Because accounts of unusual atmospheric phenomena and weather are found in contemporary historical records from Europe (Stothers, 1998), and European tree ring records show a pronounced growth anomaly (Zielinski et al., 1995;Sigl et al., 2015), the hydrological change at Store Mosse may well have been induced by this eruption. An effect of tephra deposition is, however, unlikely, because the nature of, and the distance from, the eruption imply that only very few tephra particles may have reached the bog. The lack of evidence of widespread wet shifts in NW European bogs at the time also makes it is difficult to link UHTZ to a change in atmospheric circulation induced by stratospheric input of sulphate aerosols. On the other hand, acid deposition can be rather localized, and the strong representation of E. vaginatum within the lower part of UHTZ is consistent with the vegetation response in field experiments (Payne and Blackford, 2005). Interestingly, the pollen diagram presented by Svensson (1988b) for a coring site within a few hundred metres of SM2009 and >800 m from the bog margin, at the same peat-stratigraphic position as UHTZ, reflects a major shift in forest composition in the surrounding landscape. Both Fagus and Picea expand abruptly and become important forest components, and Pinus begins a gradual expansion. Because these taxa are all associated with acidic soils, this marked pollen-stratigraphic change could reflect vegetation response to an acid deposition event, but there are certainly alternative explanations. Still, it is reasonable to attribute UHTZ to an impact of the synchronous Eldjgá eruption, with acid deposition as the main candidate.
Implications for the origin of wet shifts in NW European ombrotrophic bogs As noted above, the strong similarity between the BSW records across LHTZ and UHTZ suggest that the Store Mosse bog system features a fundamental set of interacting processes enabling consistent registration of BSW changes despite large differences in bog community composition and climatic conditions. Although Store Mosse is unusually large, most ombrotrophic bogs in southern Sweden show comparable vegetation and microtopography and are therefore likely to have registered wet shifts through operation by the same processes, in particular those wet shifts associated with abrupt transitions from well-decomposed to well-preserved Sphagnum peat known as recurrence surfaces (Granlund, 1932). It is less straightforward to apply the conceptual model of BSW registration inferred from the Store Mosse records to ombrotrophic bogs in other parts of NW Europe, but many wet shifts reported from this region may have been registered in a similar way, especially those characterized by conspicuous layers of well-decomposed peat.

Conclusions
Based on the data from the Store Mosse sequence we conclude the following: • Internal feedback processes involving vegetation composition, microtopography and degree of peat decomposition contributed to the large amplitude and rapid rate of BSW change registered for the two wet shifts dated to c. 2700 and 1000 cal a BP, and the bog system was able to operate in a consistent way despite different internal and external conditions. • Both wet shifts are attributed to events of external forcing (a change in atmospheric circulation induced by a solaractivity decrease and acid deposition from a volcanic eruption), but their large amplitude is likely to partly be an effect of high bog-system sensitivity at the time of forcing developed in response to previously very dry conditions. • Ombrotrophic bog systems comparable to Store Mosse are expected to register BSW changes through operation by the same processes, but regional registration of major wets shift may require that bog systems previously have developed a common high hydrological sensitivity, e.g. in response to a change in atmospheric circulation. • Events of solar and volcanic activity are likely to have induced wet shifts in NW European bogs during the Holocene.