Late survival of woodland contrasts with rapid limnological changes following settlement at Kalmanstjörn, Mývatnssveit, northeast Iceland

The settlement of Iceland is known to have had profound impacts on vegetation and landscape stability, but there remain uncertainties around the spatial variability and timing of environmental change, and the impacts of settlement on aquatic ecosystems. Here a new multiproxy palaeoenvironmental reconstruction spanning the last 3000 years is presented from Kalmanstjörn, a small lake in Mývatnssveit, northeast Iceland. Sedimentology, pollen and non‐pollen palynomorphs, and geochemical proxies, dated using tephrochronology, are used to reconstruct terrestrial vegetation, landscape stability and aquatic ecosystems. The data reveal complex environmental dynamics after settlement. At this site, substantial tree populations persisted until the late 15th century, in strong contrast to the rapid deforestation shown by almost all other records from Iceland. The eventual loss of woodland may have been caused by changes in direct human activity and the location of extensive grazing, in combination with Little Ice Age climatic cooling. The loss of woodland was accompanied by increased soil erosion. Conversely, the lake ecosystem showed an immediate response to settlement, becoming more productive for several centuries, perhaps in response to increased availability of nutrients from grazing herbivores. The late persistence of woodland in the Kalmanstjörn record adds to our understanding of the spatial variations in ecosystem responses to settlement in Iceland, while the evidence for decoupling of the aquatic and terrestrial systems suggests that palaeolimnological reconstructions focusing on aquatic ecosystem responses may be important new sources of information on the wider ecological consequences of human settlement.

The environment of Iceland has undergone significant changes since the colonization of the island by Norse settlers in the 9th century AD. Today, woodland covers just 1% of the island and 96% of soils are experiencing erosion (Arnalds 1987), but palaeoenvironmental reconstructions have revealed that before settlement woodland cover was much higher and erosion far less extensive ( Olafsd ottir et al. 2001;Olafsd ottir & Guðmundsson 2002;Streeter 2011;Streeter et al. 2015). The first widespread settlement of Iceland, known as Landn am (Old Norse for 'land taking'), occurred around AD 877 Schmid et al. 2017). Before settlement, changes in the Icelandic ecosystem appear to have been driven primarily by climatic change and, in this volcanic landscape, by the (often short-lived) ecological consequences of tephra deposition events (Eddud ottir et al. 2015(Eddud ottir et al. , 2016. Settlement saw the introduction of new anthropogenic pressures, including grazing by mammals, woodland clearance, and burning, which caused widespread ecological change McGovern et al. 2007;Streeter et al. 2015). These new pressures were imposed upon a system that continued to respond to climatic changes, the most important being the transition between the Medieval Warm Period (MWP; c. AD 780-1250) and the Little Ice Age (LIA; c. AD 1250-1900; Ogilvie et al. 2000;). Tephra deposition also continued to play a role in environmental change: not only did it continue to directly impact ecosystems as it was deposited, but it also changed the intensity and distribution of grazing pressures as farmers moved to protect their livestock (Wilcox 1959;Edwards et al. 2004). Understanding the relative contributions of each of these drivers (climate, volcanism, and anthropogenic activity) to environmental change is a major challenge for palaeoenvironmental research in Iceland.
Our current understanding suggests that the impacts of the settlement of Iceland on terrestrial vegetation were widespread but also spatially variable (Streeter et al. 2015). The majority of palaeoenvironmental reconstructions have suggested that, where woodland was widespread before settlement, it substantially decreased in extent within decades of Landn am (Þ orarinsson 1944;Einarsson 1961Einarsson , 1963Hallsd ottir 1987Hallsd ottir , 1995Riddell et al. 2018b). The majority of records showing rapid deforestation come from either mire sites, which produce pollen records from which local conditions (rather than large-scale regional trends) can be most easily reconstructed (Bunting et al. 2004), or from sites close to early farms, where early and rapid clearance would be expected. However, some lake sediment sequences (which should reflect a larger pollen source area), for example from Vatnskotsvatn and Helgutj€ orn, show some areas did lose woodland cover rapidly on a catchment scale. Over the last 15 years, this narrative has been challenged by new records that show persistence of woodland for several centuries after settlement. For example, the record from Breiðavatn, a lake in western Iceland, shows only a slight initial decrease in woodland extent immediately after Landn am, with large-scale deforestation occurring around AD 1300 ). In the M yvatnssveit region in northeast Iceland, a cluster of records similarly indicates that woodland survived for many centuries before deforestation to approximately present-day levels (small patches still persist today): the Helluvaðstj€ orn lake record shows only a slight decrease in Betula pubescens abundance immediately after Landn am, with a final reduction only after AD 1300 ); a nearby peat record from the Grӕnavatn estate similarly shows a small reduction in woodland extent at Landn am with a second, final decline after AD 1300 (Colquhoun et al. 2010); and a peat record from the Hofstaðir farm estate suggests that low-density woodland was preserved as late as AD 1790 before it was cleared, although the record probably reflects the vegetation of only a small area (Tisdall et al. 2018). Yet with only one regional-scale record (from Helluvaðstj€ orn), it remains unclear for how long widespread woodland cover persisted throughout M yvatnssveit and whether there was significant spatial variability in environmental changesa knowledge gap that this paper aims to address.
The impact of settlement on Iceland's freshwater aquatic ecosystems is perhaps even more poorly understood. Lake ecosystems may prove to be more sensitive to settlement and subsequent disturbances than may be inferred from terrestrial pollen data alone; for example, small-scale changes in land use (such as limited woodland clearance) may be obscured within the pollen spectra but may have a large impact on the inputs of material into a lake (Paus 2013;Stivrins et al. 2019). Therefore, the incorporation of information from the lake ecosystem can complement data obtained about terrestrial environments by providing insights into landscape dynamics on a well-defined catchment scale, particularly in small sub-arctic lakes, which are likely to be sensitive to changes in inputs (Dubois et al. 2018). Across the world, the response of aquatic systems to anthropogenic activity is typically characterized by eutrophication, often driven by the mobilization of soil nutrients into lakes and rivers after deforestation (e.g. Leahy et al. 2005;Meyers 2006;Mills et al. 2017). In Iceland, most studies using palaeolimnological proxies such as chironomids have focused on palaeoclimatic reconstructions and have thus been intentionally located far from centres of human activity (Axford et al. 2009;Holmes et al. 2016). Only two records exist that combine high-resolution palynological and palaeolimnological data through the settlement period: Helluvaðstj€ orn and Breiðavatn, both of which show only a small change in woodland extent at Landn am Gathorne-Hardy et al. 2009). At Helluvaðstj€ orn, Landn am is marked by increased concentrations of chironomid head capsules and by increased accumulation rates of aquatic pollen and spores and Pediastrum colonies; this suggests that nutrient availability increased immediately after Landn am . At Breiðavatn, Pediastrum abundance greatly increased after Landn am, again indicating a change in nutrient availability. The current paucity of data for aquatic ecosystems and for soil dynamics after settlement makes it difficult to reconstruct the nature and the causes of aquatic system dynamics in Iceland, and their causal relationships with terrestrial vegetation.
This study presents a new palaeoenvironmental data set from Kalmanstj€ orn, a lake in the M yvatnssveit area, northeast Iceland. M yvatnssveit is one of the most important areas in Iceland for understanding the dynamics of Landn am, as it has been extensively excavated and data from the region have been used to inform interpretations of settlement processes across Iceland (see below; V esteinsson & McGovern 2012). However, the region still lacks palaeoenvironmental data on a similar scale to the archaeological data set. This hinders the development of models of landscape change that are able to sufficiently account for the spatial variability that might be expected. At present, only one lake record (Helluvaðstj€ orn) is available for understanding regional-scale environmental responses to settlement in the region ( Fig. 1B; Lawson et al. 2007), but in order to improve our knowledge of the impacts of settlement dynamics on ecosystems more regional-scale lake records are required. This study uses sedimentology, organic geochemistry and palaeoecological data from a sediment record obtained from Kalmanstj€ orn, a lake in the northwest of M yvatnssveit, to reconstruct terrestrial vegetation, landscape stability, and aquatic ecosystem developments from 950 BC (approximately the time of deposition of the Hekla 3 tephra layer) to the 19th century AD. In doing so, this study aims to: (i) further develop the terrestrial palaeoenvironmental history of Iceland and the M yvatnssveit region, (ii) investigate the evolution of the aquatic ecosystem through the settlement period, and (iii) investigate the causal relationships between environmental systems in response to human activity, climate change, and tephra deposition.

Study area
M yvatnssveit is an inland area centred on Lake M yvatn, the third-largest freshwater lake in Iceland. The climate of M yvatnssveit is more continental than other areas of Iceland, with an average annual temperature range of 13-15°C, and it is among the driest areas, with average annual precipitation of 400 mm a À1 (Einarsson 1979). Lakes are covered by ice for much of the year; Lake M yvatn experiences an average of c. 190 days of ice cover per year (Rist 1979). The vegetation of M yvatnssveit is presently dominated by dwarf-shrub (Betula nana, Empetrum nigrum and Salix) and mire (Poaceae and Cyperaceae) communities in the rangelands, and mountain birch (B. pubescens ssp. tortuosa) is present in small populations where it is protected from grazing, such as on rough lava fields (Lawson et al. 2009). M yvatnssveit has one of the densest concentrations of excavated archaeological sites in the Norse North Atlantic region V esteinsson & McGovern 2012). Lake M yvatn is a relative hotspot of biological diversity, and lake resources may have provided settlers with the opportunity to supplement their pastoral subsistence strategy with fishing and eggcollecting while the earliest farms were being established and increasing in productivity (although there is little evidence that any farm adopted fishing as the dominant part of its resource strategy; V esteinsson & McGovern 2012). Archaeological data suggest that all good-quality land was probably occupied by AD 950 (V esteinsson & McGovern 2012). A series of abandoned farms in the colder, drier highlands to the south of M yvatn are thought to have been established in a second wave of settlement and survived only until the 12th century; these settlements appear to have been unsustainable in the long-term, in part due to environmental pressures such as soil erosion and climatic change in these marginal settings and in part due to societal pressures (V esteinsson et al. 2014). There is evidence across the region for charcoal production to facilitate metallurgy until the 12th century, which would require the (probably intentional) preservation of woodland populations far in excess of current B. pubescens stands to supply sufficient wood .
Kalmanstj€ orn (latitude 65°39.46'N, longitude 17°14.03'W; Fig. 1C) is a small (~0.27 km 2 ) and shallow (maximum depth <2 m) lake situated 265 m above sea level. The lake probably occupies a glacially scoured depression. The lake catchment is gently sloped and is characterized by andesitic soils overlying glacial till, with abundant small-scale (<1 m), low-relief, frostheave hummocks (th ufur). Heathland species, dominated by Empetrum nigrum, characterize the drier hummocks while members of the Cyperaceae family populate the hollows. Kalmanstj€ orn is situated on a narrow (~6 km wide) ridge between two river valleys (Lax adalur and Reykjadalur) and sits close to the watershed separating the two valleys, and to the south of Kalmanstj€ orn is the much larger Lake M asvatn (Fig. 1C). As a result, Kalmanstj€ orn has a relatively small catchment and may be expected to be disproportionately dominated by aeolian rather than hydrological inputs. As a small lake, Kalmanstj€ orn probably recruits local pollen from no further than 1-2.5 km from the lake (based on estimates from slightly smaller lakes in southern Scandinavia; Nielsen & Sugita 2005;Hellman et al. 2009;Hjelle & Sugita 2012). The upper end of this range suggests that the local Kalmanstj€ orn signal will broadly reflect the vegetation dynamics of the ridge on which Kalmanstj€ orn lies, while the remainder of the assemblage (potentially over 50% of the pollen signal; Nielsen & Sugita 2005) will be representative of the vegetation of the wider region.
Today, three farms are operational within 3 km of the lake: M askot (occupied from at least the 16th century; V esteinsson 2011), Hr ısar and Br un (Fig. 1C). Archaeological evidence indicates that the local landscape on the edge of the ridge has been occupied since shortly after Landn am. Þyrilskot (Fig. 1C), which was probably never fully established as a farm, preserves evidence for soil improvement prior to AD 940, with a small enclosure built after AD 940 (V esteinsson & McGovern 2012). Viðat oft, on the southwest corner of M asvatn (Fig. 1B), comprises two concentric enclosures, the innermost of which currently encloses a wet bog; the enclosures may represent attempts to protect an early area of intensive productivity in the inner enclosure and a later, more expansive, area of cultivation when the attempts at intensive cultivation proved unproductive (V esteinsson 2011). The Hallskot farm ( . Wall structures pre-dating the 12th century are prevalent in the two valleys, representing the delineation between exclusive farmland and the communal rangelands (Einarsson 2015). Across the region there is evidence for abandonment of farms around the 12th century, reflecting a social restructuring, which involved a concentration of households onto a smaller number of farms; this is likely to have affected the land use of outfields, but little palaeoenvironmental evidence is currently available for the impacts that this had on ecosystems (V esteinsson 2011). There was, therefore, probably a consistent local land use for the first three centuries of the settlement period, possibly followed by a change in land use from the 13th century onwards. The modern and archaeological evidence suggest that the Kalmanstj€ orn catchment has remained on the margins of areas of more intensive land management since settlement.

Sampling and sedimentology
A 290-cm sediment core was obtained from the centre of Kalmanstj€ orn using a Russian corer with a 50-cm chamber. Core segments were obtained with a 10-cm overlap, and segments were correlated using stratigraphical information (e.g. tephra layers) and magnetic susceptibility measurements (see below). The core was stored at 4°C until analysis. The sediment composition was described using the Troels-Smith method (Troels- Smith 1955). The uppermost 10 cm of the core was not subsampled for analysis as the integrityof the sediment in this section was poor. Volumetric magnetic susceptibility was measured on 1-cm 3 subsamples at 2-cm intervals using a Bartington Instruments MS2 meter and a Bartington MS2G probe. 1-cm 3 subsamples were taken at 2-cm intervals for loss-on-ignition analysis; the subsamples were dried at 105°C for at least 12 h for dry bulk density determination and ignited at 550°C for 4 h for organic matter determination (Heiri et al. 2001;Santisteban et al. 2004).

Geochemical analysis
Samples for geochemical analyses were taken at depths that correspond with samples taken for pollen analysis (see below). Samples were oven-dried at 40°C, then crushed into a fine powder and homogenized using a pestle and mortar (Nieuwenhuize et al. 1994). 20 mg of material was weighed into a capsule and acidified with 10% HCl to remove inorganic carbon. Samples were analysed using an Elementar vario EL cube CHNS Elemental Analyser using sulfanilamide as a standard. The atomic carbon and nitrogen concentrations were used for calculation of the C:N ratio.

Pollen, spore and Pediastrum analysis
Samples for pollen analysis were taken at 8-cm intervals, increasing to 4-cm intervals above the Landn am tephra layer and 2-cm intervals around tephra layers and around periods of palaeoecological change. Pollen preparation followed standard procedures using a density separation technique (Nakagawa et al. 1998;Campbell et al. 2016), as follows. 1-cm 3 subsamples were spiked with Lycopodium clavatum tablets to allow estimates of concentration (Stockmarr 1971). Samples were treated with 10% HCl and 10% KOH, then sieved to remove material >180 lm. Lithium heteropolytungstate (tradename LST Fastfloat) at a density of 1.88 g cm À3 was used for density separation. Samples then underwent acetolysis, then dehydration using tert-butanol. Samples were stored and mounted in silicone oil.
Pollen and non-pollen palynomorphs were identified at 4009 magnification (with critical examination where required at 10009 magnification) until at least 300 terrestrial pollen grains had been counted (cf. Maher 1972). Some Icelandic studies exclude Cyperaceae from the pollen sum where it dominates an assemblage to avoid over-representation of a local signal (e.g. G ıslad ottir et al. 2010; Riddell et al. 2018a). However, because Kalmanstj€ orn has an area of~0.27 km 2 and the core was taken from the centre of the lake (approximately 200 m from the nearest shore) the Kalmanstj€ orn record was not expected to be (and did not appear to be) disproportionately affected by local Cyperaceae populations around the margin of the lake, and Cyperaceae was, therefore, included in the pollen sum. Pollen identification followed Moore et al. (1991) and the pollen reference collection held at the University of St Andrews, and taxonomy and nomenclature follows Lawson et al. (2007). Betula pubescens and B. nana produce morphologically similar pollen. Typically the grains of B. pubescens are larger than those of B. nana, although hybridization between the two species can produce intermediate grain sizes and there is often an overlap in the size distributions (Karlsd ottir et al. 2007). Nevertheless, separation of the two species using grain diameter can be used to generate an estimate of the proportion of B. pubescens and B. nana grains in a sample. A number of separation thresholds have been proposed, but here we follow earlier separation criteria: Betula grains were measured at 10009 magnification, and those <20.5 lm in diameter were assigned to B. nana while all others were assigned to B. pubescens (M€ akel€ a 1996; Caseldine 2001). Each grain assigned to B. pubescens was assessed for evidence of damage and recorded in four categories: corroded, degraded, folded/crumpled, and broken (Elsik 1966;Cushing 1967;Havinga 1967). Pollen data were zoned based on terrestrial pollen using the binary splitting sum-of-squares method in Psimpoll (Bennett 2009). The record was separated into two statistically significant zones that are divided at c. AD 1480. To aid the following description and interpretation of the data, the earliest zone is separated into two subzones representing the pre-settlement (KMN-1a) and post-settlement (KMN-1b) periods, separated by the Landn am tephra layer (see below for identification and correlation).
Identification of coprophilous fungal spores followed Feeser & O'Connell (2009) and Cugny et al. (2010). Pediastrum colonies were counted where more than half of the colony was intact, and identification followed Jankovsk a & Kom arek (1995,2000) and Kom arek & Jankovsk a (2001). Proportions of terrestrial spores, aquatic pollen and spores, fungal spores and Pediastrum colonies were calculated using the sums of the terrestrial pollen plus all taxa in the relevant group. Microscopic charcoal fragments >10 lm were identified but abundances were consistently low with no discernible patterns, and data are, therefore, not presented here.

Tephra geochemistry and age model
Tephra layers were selected for geochemical analysis based on their stratigraphical integrity, thickness and discreteness, focusing on parts of the stratigraphy where important changes in the proxies occurred. In total, eight layers were selected for analysis ( Fig. 2; Table S1), while two additional layers were excluded as they were diffuse and are, therefore, unlikely to represent primary tephra deposition. Tephra shards were isolated from the sediment matrix using standard acid digestion procedures whereby samples were treated with hot concentrated H 2 SO 4 and HNO 3 (Dugmore et al. 1992). Shards were mounted in an epoxy resin (tradename EpoThin2), ground to expose shard surfaces, polished, and carbon coated (Hall & Hayward 2014). Tephra samples were analysed using a Cameca SX100 electron microprobe at the University of Edinburgh for the determination of major elemental oxides. Analysis used an accelerating voltage of 15 kV, a current of 2 nA/80 nA, and a beam diameter of 8 lm in order to avoid the mobilization of Na, and subsequent overestimation of Al and Si and underestimation of Na and K (Hayward 2012). Analytical accuracy and precision were determined using measurements of a rhyolitic (Lipari) and basaltic (BCR2g) standard at the beginning of a set of analyses and then every 24 h thereafter (Table S2).
Correlation of tephra layers was based on comparison with published geochemical data (Tephrabase: http://www.tephrabase.org; Larsen et al. 2002;Gudmundsd ottir et al. 2012) and with reference to other records with tephrostratigraphical profiles in the region (Larsen et al. 2002;Lawson et al. 2007;Adderley et al. 2008;Brown et al. 2012;Gudmundsd ottir et al. 2012;Sigurgeirsson et al. 2013;Barclay 2016;Tisdall et al. 2018). Six tephra layers in the record have been correlated with independently dated tephras (Fig. 2): Hekla 3 at 152 cm (radiocarbon dated to 2879AE34 14 C a BP: Dugmore et al. 1995); the base of the Landn am tephra layer (LTL) at 118 cm (dated to AD 877AE1: Schmid et al. 2017); the base of the V-Sv tephra layer at 106.5 cm (previously referred to as V-950 and now dated to AD 938AE6: Schmid et al. 2017); the base of a Veiðiv€ otn layer at 85.5 cm that is geochemically and stratigraphically similar to one found in nearby Lake M yvatn (dated to AD 1159: Sigurgeirsson et al. 2013); the base of V-1477 at 44 cm (historical records providing a date of AD 1477); and the base of V-1717 at 27 cm (historical records providing a date of AD 1717). Using these six tephra layers and the year of coring as the age for the top of the core, an age model was constructed using a Bayesian approach in the Bacon package in R version 3.4.0 ( Fig. 3; Blaauw & Christen 2011; R Core Team 2017). The Bacon age-depth model normally permits only small variations in sediment accumulation rate across dated horizons in order to best simulate a natural depositional environment, but due to the significant change in accumulation rate observed across the LTL (which has a precedent in the Helluvaðstj€ orn record; Lawson et al. 2007), a boundary was included in the model at the depth of the LTL to permit a larger change in sediment accumulation rate (Fig. 3).

Sedimentology and organic geochemistry
Magnetic susceptibility (MS), dry bulk density (DBD) and organic matter (OM) covary through the succession (Fig. 4). Typically, high OM values correspond with low DBD and MS values (and vice versa). OM shows the most inter-sample variability. The most prominent long-term trends include a slow reduction in OM above the Hekla 3 tephra layer and an increase in OM above the LTL. Above the LTL, high OM values are interrupted only where tephra layers are found. The largest response, seen in all sediment properties, is to the V-1477 tephra. Sediment accumulation rates show a marked increase across the LTL horizon, increasing from 0.02AE0.001 cm a À1 below the LTL to an average of 0.13AE0.08 cm a À1 above the LTL (Fig. 3).
Trends in total organic carbon (TOC) and total nitrogen (TN) closely follow OM trends: they decline just beneath the LTL, then increase above the LTL with sharp, short-lived decreases coinciding with tephra layers (most prominently the V-1477 tephra layer). C:N ratios through the record are typically <15, with sharp increases coinciding with tephra layers. Between tephra layers there is some evidence of C:N values gradually increasing (for example between the V-Sv and V-1159 layers, and between the V-1159 and V-1477 layers) before returning to previous values above the upper tephra layer (Fig. 4).
Terrestrial spores are sparse until AD 650 when Selaginella selaginoides, Lycopodium annotinum and Sphagnum percentages increase (Fig. 5B). Of these S. selaginoides is the most abundant, reaching a maximum of 5%. Aquatic pollen and spores are dominated by Myriophyllum alterniflorum, which varies between 10 and 30%. Higher values are attained around AD 70 and AD 870. Isoetes consistently forms a minor component (<2%) of the aquatic assemblage. The number of Pediastrum colonies increases gradually before stabilizing at maximum values after AD 700. Geochemical tephra biplots. Data from the Kalmanstj€ orn record are overlain on data for the source volcanic system and the eruption to which the layer has been correlated (data from Hannon et al. 1998;Boygle 1999;Larsen et al. 1999Larsen et al. , 2002Wasteg ard et al. 2001;Wasteg ard 2002;Pilcher et al. 2005;Lawson et al. 2007;Gudmundsd ottir et al. 2012;Sigurgeirsson et al. 2013;Streeter & Dugmore 2014). Raw data can be found in Table S1.
Zone KMN-1b (AD 877-AD 1477). -Zone KMN-1b represents the first 600 years following settlement and is, like KMN-1a, characterized by terrestrial pollen assemblages dominated by arboreal pollen that show little variability except in samples around tephra layers (Fig. 5A). B. pubescens percentages vary between 35 and 45%, with lower values coinciding with tephra layers. B. nana (10-15%, with a large short-lived peak to 20% at AD 1090), Poaceae (7.5-15%) and Cyperaceae (25-35%) are similarly stable, with their highest values occurring when tephra layers are present. The tephra layers that coincide with the largest assemblage changes are the LTL and the V-1159 layers. The woody shrub taxa (Salix and Juniperus communis) maintain their low percentages from KMN-1a. J. communis declines to consistently low (<2%) percentages after AD 1300. Ericaceae remain scarce and decrease to their lowest values in the record after AD 1200. Minor taxa (including Thalictrum and members of the Apiaceae and Ranunculaceae families) increase slightly above the LTL.
Terrestrial spores retain their low values from KMN-1a with the exception of AD 1110-1170, when S. selaginoides percentages increase slightly (Fig. 5B). The coprophilous spore types Sporormiella and Sordaria appear sporadically at low percentages after AD 950. Percentages of aquatic pollen and spores increase. Isoetes percentages increase to vary between 2 and 4% immediately above the LTL. M. alterniflorum percentages initially decrease until AD 950. They then increase to >30% between AD 1100 and AD 1350, before gradually decreasing through the remainder of the zone (with the exception of a peak to >30% at c. AD 1400). The main Pediastrum taxa (Pediastrum boryanum undiff., P. integrum, and P. subgranulatum) that appeared in KMN-1a maintain percentages similar to the previous zone. P. boryanum var. cornutum becomes a significant part of the assemblage during KMN-1b and exhibits two large (>35%) peaks at AD 1050 and AD 1400 with values >10% between the peaks. arboreal pollen and increased dominance of heathland and herbaceous taxa (Fig. 5A). B. pubescens percentages decline from 40 to 15% over approximately 10 years at the base of the zone and then show little variability. This reduction is accompanied by a sharp increase in Cyperaceae, from 25 to 45%, over the same time scale. B. nana, Empetrum nigrum and Thalictrum also show small-scale percentage increases at this time.
Poaceae retains similar percentages to KMN-1. Salix maintains values from KMN-1b (except for a sharp, short-lived decrease to <1% at AD 1750), and Juniperus communis appears sparsely after AD 1580. Members of the Apiaceae family become rare after AD 1480. Terrestrial spores increase in importance in KMN-2 (Fig. 5B). Lycopodium annotinum and Selaginella selaginoides both immediately increase to >5% and remain at these levels for the rest of the zone. Sphagnum shows a slight increase at the same time. Continued minor Sordaria occurrences, and a single occurrence of HdV-172 (Coniochaeta lignaria), represent the coprophilous fungal spore assemblage of the zone. Aquatic species decline rapidly above the V-1477 tephra layer: of the two most abundant aquatic taxa, Isoetes returns to KMN-1a levels and Myriophyllum alterniflorum declines to its minimum values in the record (<2%). Both aquatic taxa show a sharp increase after AD 1750. Figure 6 shows summary data for concentrations and accumulation rates of the three main categories of taxa (terrestrial pollen, aquatic pollen and spores, and Pediastrum colonies) and the dominant taxon in each category (Betula pubescens, Myriophyllum alterniflorum, and Pediastrum boryanum undiff., respectively). Terrestrial pollen concentrations follow the same broad trends as the percentage data and are relatively consistent except for sharp decreases where tephra layers occur. Accumulation rates of terrestrial pollen increase from <150 grains cm À2 a À1 in KMN-1a to >500 grains cm À2 a À1 in KMN-1b, before declining in KMN-2.

Concentrations and accumulation rates
Concentrations and accumulation rates of aquatic pollen and spores closely follow the M. alterniflorum percentage trends; low values dominate in KMN-1a, higher values in KMN-1b, and low values in KMN-2. Pediastrum colony data, on the other hand, show a permanent increase from the base of KMN-1b. Low values in KMN-1a (<4000 colonies cm À3 and <100 colonies cm À2 a À1 ) are followed by markedly higher concentrations (5000-16 000 colonies cm À3 and 500-2000 colonies cm À2 a À1 ) in KMN-1b, with the lowest values coinciding with tephra layers. Concentrations maintain these values in KMN-2, whereas accumulation rates drop to~300 colonies cm À2 a À1 .

Pollen preservation
The Betula pubescens pollen preservation data show two distinct periods (Fig. 7). Below the V-1477 tephra layer (in KMN-1a and KMN-1b), on average less than 20% of B. pubescens grains are damaged. Inter-sample variability is noticeablygreater above the LTL (in KMN-1b) than below it, but this is partly a consequence of higher sampling resolution and increased sedimentation rates. Instances of increased damage frequently coincide with tephra deposition. Above V-1477 (in KMN-2) the proportion of damaged grains increases permanently to 40-50%. Throughout the record (and most prominently in KMN-2) the most common types of damage to B. pubescens grains are folding/crumpling and degradation.

Terrestrial vegetation
Pre-settlement environment (850 BC-AD 877). -The pre-settlement environment is characterized by stable pollen spectra dominated by high proportions of Betula pubescens and with very little variability in any taxa. B. pubescens pollen consistently exceeds 40% of the assemblage, which could be interpreted to represent widespread woodland in the vicinity of the lake. However, the B. pubescens pollen accumulation rate is low by comparison with measurements of modern pollen accumulation rates in Scandinavia, suggesting that B. pubescens populations at Kalmanstj€ orn might in fact have been more sparse and/or distant than would be inferred from the pollen percentage alone. Lake surface sampling and pollen trap data have been used to identify palynological influx values that can signal the local presence of Betula populations in Scandinavia. For example, a presence/absence threshold of 250 grains cm À2 a À1 has been identified in core-top sediments (Sepp€ a & Hicks 2006), while data from pollen traps suggest that a site within a dense birch woodland could exceed accumulation rates of 1500 grains cm À2 a À1 (Hicks & Hyv€ arinen 1999). However, there is some evidence that suggests that Betula accumulation rates in Iceland are consistently lower than in Scandinavia (H€ attestrand et al. 2008). During the pre-settlement period in the Kalmanstj€ orn record, B. pubescens accumulation rates never exceed 100 grains cm À2 a À1 , and the combined accumulation rates of B. pubescens and B. nana exceed 100 grains cm À2 a À1 in only one sample. In this, the Kalmanstj€ orn record is similar to the Helluvaðstj€ orn record, in which pre-settlement accumulation rates of B. pubescens likewise fall consistently below 100 grains cm À2 a À1 ).
There is a large amount of independent evidence that woodland was widespread in M yvatnssveit immediately after settlement when resource utilization can be detected in records, including charcoal pits, charcoal lenses in archaeological contexts, and macrofossil assemblages (Lawson et al. 2009), at a time in which the Kalmanstj€ orn record continues to record high proportions of B. pubescens pollen (see below). Given that the pre-settlement pollen assemblages so closely compare with the post-settlement assemblages in which there is independent evidence for local and regional woodland populations, and the evidence that local Betula populations in Iceland can produce markedly less pollen than Scandinavian populations (H€ attestrand et al. 2008), we interpret the Kalmanstj€ orn record to show that the pre-settlement environment was characterized by widespread woodland populations on both a local and regional scale. Between c. AD 100 and Landn am (AD 877) the proportion of B. pubescens grains showing evidence of corrosion and degradation increases. Pollen that has been deposited and stored in soils typically shows signs of chemical damage caused by bacterial and fungal attack, and grains that are subsequently remobilized may suffer mechanical damage (Cushing 1964(Cushing , 1967Havinga 1967). Therefore, the preservation data indicate that there was increased reworking of soils containing non-contemporaneous pollen grains into the lake. It is most likely that woodland in the hydrological catchment contracted and exposed soils to erosional processes. This could perhaps have been caused by increased wetness of soils creating unfavourable conditions for B. pubescens growth (Bartley 1973). Alternatively, woodland contraction and increased mobility of soils may have been caused by cooling climatic conditions. Increased soil mobilization beginning around AD 500 is a feature of a composite record from a number of lake records in western Iceland, coinciding with evidence for increasingly cold and harsh climatic conditions (Geirsd ottir et al. 2020). This correlates with a composite of 25 Betula undiff. records that shows a gradual decline in Betula pollen abundance starting around 100 BC, suggesting that these largescale environmental conditions were affecting vegetation across much of the island (Streeter et al. 2015).
Post-settlement preservation of woodland (AD 877-AD 1477). -In the context of the existing palaeoenvironmental data from M yvatnssveit and from the wider Icelandic context, which for the most part show deforestation by AD 1300, the most intriguing element of the Kalmanstj€ orn record is the persistence of abundant B. pubescens pollen until AD 1480, nearly 600 years after the area was first settled. The pollen assemblages show remarkable stability across the Landn am horizon, and throughout KMN-1a and KMN-1b the B. pubescens pollen counts consistently fall within the statistical uncertainty between samples (Fig. 7). The only exceptions to the stability of the pollen record are what appear to be low-magnitude and short-lived responses to the deposition of the V-Sv (AD 938AE6) and V-1159 tephra layers where counts lie outside of inter-sample variability. These responses probably reflect changes in substrate conditions that led to increased growth rates and pollen productivity of herbaceous taxa, thereby reducing the relative abundance of B. pubescens pollen (Edwards et al. 2004;Payne et al. 2013). As this overall stability is unusual in post-settlement Icelandic pollen records, it is important to establish whether the pollen assemblages (and, in particular, the B. pubescens record) could be reflecting any processes other than the persistence of woodland. Firstly, the climatic developments of the MWP and LIA may have affected pollen productivity. The MWP was a period of increased average temperatures in the North Atlantic region that may have started around AD 800 in Iceland, while the LIA was a cooler, drier and windier period that began between AD 1250 and AD 1300, with intensification after AD 1400 or AD 1450 ( has been positively correlated with temperature (Hallsd ottir 1999). Therefore, increased pollen productivity during the MWP could maintain high B. pubescens proportions in pollen assemblages even as trees were being felled (cf. ). However, this is unlikely for two reasons. There is no sign of an increase in B. pubescens in any of the three data forms (proportion,concentration,and accumulation rate;Figs 5,6) in the period of the MWP just before settlement (AD 800-877). There is also no sign of a reduction in B. pubescens in any of the three data forms after AD 1250 when the climate began to cool, which suggests that B. pubescens percentages have not been substantially affected by climatic conditions. Secondly, high B. pubescens percentages in the record may be artificially increased by the influx of non-contemporaneous grains delivered to the lake by the reworking of soils (particularly in a landscape denuded of trees). However, proportions of damaged B. pubescens grains remain low and C:N ratios remain stable, indicating that there were no prolonged periods in which there was increased delivery of terrestrial material into the lake (Meyers & Ishiwatari 1993;Meyers 2006). Therefore, it is unlikely that reworking of old pollen grains has substantially altered the pollen assemblages between settlement and AD 1480. The pollen abundance data, therefore, strongly indicate that there was long-term persistence of woodland near Kalmanstj€ orn between the Landn am (AD 877AE1) and V-1477 (AD 1477) tephra layers. Whether this was an accidental outcome of land-use decisions, or whether it represents an intentional preservation effort, can be assessed using the available data. There are sporadic occurrences of coprophilous fungal spores (Sordaria and Sporormiella) after Landn am, which may indicate that herbivores were allowed to freely move in the nearby catchment (Fig. 8). Coprophilous fungal spores are not widely dispersed due to their production near the ground, and therefore, they can be taken to represent herbivore presence very close to the lake (Van Geel et al. 2003). Given the limited evidence for landscape instability that could have transported more distal spores into the lake, the coprophilous fungal spore record can be taken to indicate a local introduction of herbivores on at least a limited scale, which would be expected to have a negative impact on woodland populations (e.g. Arad ottir & Arnalds 2001). However, other evidence suggests that woodland was intentionally protected. There are abundant charcoal production pits near the lake and in the region as a whole, dating late into the settlement period, which indicates that people were using the woodland for resources in a long-term, sustainable manner McGovern et al. 2007;Lawson et al. 2009). Additionally, throughout the region there are the remains of wall structures that were in use between the 10th and 12th centuries in part constructed to separate exclusive land from the communal grazing lands (affret ır) (Einarsson 2015). Such wall structures are extensive in the Lax adalur and Rekyjadalur valleys and indicate that efforts were being made to manage the use of the landscape between the two rivers (the Kal-manstj€ orn catchment), which, based on the archaeological record, would have been of interest to a large number of different farms (V esteinsson & McGovern 2012;Einarsson 2015). Finally, despite the fungal spore occurrences, grazing tolerant taxa that would indicate the presence of herbivores remain rare. It is likely that the vegetation signal is more robust than the coprophilous fungal spore data, which are not always a clear proxy for herbivore abundance (Gill et al. 2013;Tisdall et al. 2018;Davies 2019). On the balance of evidence, it is, therefore, most probable that efforts were made to protect the local woodland resources, with perhaps only low-density grazing occurring in the area surrounding Kalmanstj€ orn.
Significance of long-term preservation of woodland in M yvatnssveit and Iceland. -The long-term preservation of woodland indicated by the Kalmanstj€ orn record supports other records from the M yvatnssveit region. The current understanding of the vegetation history of M yvatnssveit is that, on the eve of settlement, some areas had extensive woodland cover whereas other areas (particularly meadows and bogs; cf. Lawson et al. 2009;Barclay 2016) were largely unwooded. After settlement, these patterns continued largely unchanged, with woodland persisting for centuries. The Hel-luvaðstj€ orn lake record and a peat profile from the Grӕnavatn estate indicate that woodland extent slightly decreased just after Landn am but was largely preserved until AD 1300, when it dropped to present levels Colquhoun et al. 2010). A peat profile from the Hofstaðir estate indicates that Betula populations survived locally until AD 1790 (Tisdall et al. 2018). The Kalmanstj€ orn record contributes to a growing story in which deforestation around Lake M yvatn was temporally and spatially variable. The area was settled rapidly (V esteinsson & McGovern 2012), and yet rapid clearance of land is not seen in any palaeoecological record from the region (although there are few chronologically robust records with high temporal resolution). It is likely that the pre-settlement environment was patchy, and those areas that were already largely free of woodland (see the meadows of Gautl€ ond and Viðatoft, and bogs such as Hr ısheimar; Fig. 1) were preferentially targeted for the first settlements (V esteinsson 1998;Lawson et al. 2009;Barclay 2016). After settlement, efforts were made to use woodland resources sustainably and woodland was preserved initially on a large scale (represented by the Helluvaðstj€ orn and Kalmanstj€ orn records) and, when those resources were depleted, on local (farmstead) scales (represented by the Hofstaðir record).
Such persistence of woodland after settlement is a rare scenario in other parts of Iceland. The only lake record outside of M yvatnssveit that indicates longterm woodland preservation is Breiðavatn in west Iceland, where woodland persisted until AD 1200 ). Additionally, a number area of charcoal production pits in the Eyjaffjallahreppur area in the south of Iceland have been dated to two distinct time periods: from Landn am to the 11th century, followed by a hiatus, then from the late 12th to late 13th centuries ). This indicates that woodland must have survived in this region until at least the late 13th century. Most other regions in Iceland preserve evidence for rapid loss of woodland. Early research prioritized sites close to early farms, which are more likely to show early localized woodland clearance (Þ orarinsson 1944;Einarsson 1961Einarsson , 1963. More recent studies support the conclusion of rapid deforestation in at least some locations, including in the Mosfell valley in the southwest (P etursd ottir 2014;Riddell 2014;Riddell et al. 2018b) and Helgutjorn in the east (J onsson 2009). A lake record from Vatnskotsvatn in the north shows deforestation before AD 950, which indicates rapid catchment-scale deforestation (Hallsd ottir 1995). Whether long-term and large-scale preservation of woodland is, therefore, a story unique to the M yvatnssveit region remains to be seen as further studies are carried out in other parts of Iceland.
Fifteenth-century deforestation (AD 1477-AD 1850). -It is not until after the deposition of the V-1477 tephra layer that the Kalmanstj€ orn record shows evidence of pronounced terrestrial change. In as little as 10 years, the woodland area contracted and was replaced by vegetation similar to that found in the lake catchment today: sedge communities around the lake margin and heathland beyond. Given the speed of the ecosystem transition, and without widespread burning (for which there is no evidence in the charcoal record), it is probable that direct clearance was in part responsible for the deforestation. At the same time, the increase in grazing-resistant taxa (such as Selaginella selaginoides and Thalictrum) and reduction in those susceptible to grazing (particularly Angelica spp.) indicate that grazing increased in intensity (Tisdall et al. 2018;Davies 2019). Both of these lines of evidence suggest that the local land use rapidly changed from one of low-intensity use to one of higher intensity resource utilization for the first time in its 600 years of settlement history.
Such a large change in local land use suggests that there was a change in the pressures on the local farmers. One external pressure was likely the climatic change in the LIA. The Icelandic climate began to cool from  c. AD 1250, with the trend intensifying after c. AD 1450 ). Such cooling may have increased the demand for woody resources (e.g. for fuel), and the woodland around Kalmanstj€ orn would have represented a prime (and rare) source in the late 15th century. Additionally, the cooling climate (exacerbated by the deposition of the thick V-1477 tephra) led to widespread landscape degradation in the highlands to the south of Lake M yvatn ( Olafsd ottir & Guðmundsson 2002;V esteinsson et al. 2014). Degradation reduced the amount of viable land in communal grazing areas, and livestock would have needed to be relocated to suitable rangeland sites at lower altitudes (such as Kalmanstj€ orn). The increase in livestock populations would have damaged Betula seedlings, preventing the replacement of lost trees (Gill 1992;Arad ottir & Arnalds 2001). Compounding this, the cooler temperatures slowed the growth of those seedlings that survived, further slowing the replenishment of woodland stocks (Wastl et al. 2001;Weih & Karlsson 2001). The cumulative impacts of these interlinked changes ultimately led to the regime shift seen in the Kalmanstj€ orn record: a sudden collapse in woodland cover.
Previous research has suggested that deforested woodlands in Iceland were typically replaced by one of two ecosystems. Where early clearance was followed by an expansion of Poaceae it has been taken to represent the establishment of infields, where fodder and sometimes cereals were grown, and outfields, where animals were permitted to graze (Smith 1995;Dugmore et al. 2005;Sveinbjarnard ottir et al. 2007). Elsewhere woodland was replaced by heathland taxa, which has been taken to represent marginal sites where little active land management occurred . At Kalmanstj€ orn, Betula pubescens was replaced by Empetrum nigrum, Lycopodium annotinum and Selaginella selaginoides in the catchment, which we infer to indicate that open heathland environments dominated. E. nigrum and S. selaginoides are also acidophilic, suggesting that soils were in an acidic state after deforestation. Elsewhere, it has been suggested that the opening of the canopy and the resultant penetration of rainfall after deforestation contributed to leaching and soil acidification (cf. Lawson et al. 2007). At Kalmanstj€ orn, where acidification immediately followed deforestation, it is unlikely that deforestation led to acidification. Instead, acidification was probably caused by the impacts of the deposition of the V-1477 tephra. The large magnitude of this event would have caused soil acidification via acid precipitation (Payne & Blackford 2005). Unlike in the majority of post-settlement Icelandic records, Poaceae never exhibits a sustained increase in extent at Kalmanstj€ orn. A low-magnitude, short-lived increase in Poaceae coincides with the V-1477 tephra, which probably reflects a short-lived period in which the tephra created a dry substrate before it was either incorporated into the soils or eroded (Edwards et al. 2004;Payne et al. 2013). Similar short-lived increases that coincide with tephra layers occur throughout the succession. The lack of any permanent change in Poaceae pollen abundance supports other lines of evidence (above) that Kalmanstj€ orn has always represented a largely unmanaged site that was never the location for infields or outfields. Instead, Kal-manstj€ orn has primarily been used as grazing lands, with grazing intensity varying through time.

Aquatic ecosystem
At the start of the record, the Kalmanstj€ orn aquatic ecosystem was for the most part a moderately productive ecosystem. Myriophyllum alterniflorum dominated the macrophyte assemblage and remained broadly stable within a 10% range and most Pediastrum taxa were present with little variability between samples. There is indication that the system slowly began to increase in productivity from c. AD 50 onwards; M. alterniflorum increases to its maximum pre-Landn am value, albeit for only one sample, and a number of Pediastrum taxa show gradual increases in abundance. AD 50 broadly coincides with the increase in the proportion of damaged B. pubescens grains in the record and indicates that there was likely an increase in the nutrients being delivered by influxes of soils into the lake (see above; Mills et al. 2017). The fact that Pediastrum taxa show a sustained increase in abundance while the aquatic macrophyte population soon returned to lower values may indicate that the two groups have different sensitivities, and perhaps that initial influxes of nutrients that were sufficient for the enhanced growth of macrophyte populations were not sustained.
The aquatic record shows evidence of substantial and sustained change only after the deposition of the LTL (AD 877AE1), reaching the highest values of macrophyte abundance in the record after AD 1000. This marks a clear decoupling between the aquatic system and the terrestrial system, which shows limited change above the LTL. The increase in M. alterniflorum pollen, in combination with the increased organic content of the sediment, indicates that macrophyte biomass increased substantially shortly after Landn am (Edwards et al. 2000). While this may have been driven in part by the climatic amelioration of the MWP, given the timing it is quite likely that settlement processes were also involved in increasing lake productivity. The introduction of browsing mammals including goats and cattle would have provided a new route for the nutrients held in trees and shrubs to be introduced into streams and lakes. A similar pattern of increased aquatic productivity was identified at Helluvaðstj€ orn, which showed the same disconnect between the terrestrial and aquatic systems , and in other parts of Iceland J onsson 2009). On the Faroe Islands (Hannon et al. 2005;Lawson et al. 2005), a similar aquatic response to human settlement in an environment lacking large tree populations provides further support to two dominant conclusions: that the introduction of grazing animals was the dominant driver of large-scale aquatic ecosystem changes; and that aquatic environments may be more sensitive to initial human disturbances than terrestrial environments in parts of the North Atlantic island chain.
Macrophyte productivity began to decline around AD 1270, and the rate of decline increased across the V-1477 horizon (just as woodland rapidly declined). The aquatic and terrestrial systems continued to be decoupled; although there is evidence (in the pollen preservation and C:N ratio data) in the Kalmanstj€ orn record for large-scale terrestrial vegetation changes and landscape instability in response to changes in human activity after AD 1477, the aquatic ecosystem does not straightforwardly reflect these terrestrial processes. Instead of increasing in response to the influxes of nutrients from increased local herbivore populations and soils, aquatic productivity decreases, counterintuitively. The decline in aquatic productivity from c. AD 1270 coincides with the onset of the LIA (Geirsd ottir et al. 2009). The LIA was probably the coolest period of the 3000 years represented by the Kalmanstj€ orn record Blair et al. 2015;Cabedo-Sanz et al. 2016). It is, therefore, likely that cooling climatic conditions led to increased lake ice cover which would limit light penetration and the remobilization of nutrients stored in lake sediments, in turn reducing productivity (Rouse et al. 1997;Ohlendorf et al. 2000;Schmidt et al. 2008). The response of the Kal-manstj€ orn aquatic system bears similarities to other aquatic records from Iceland that cover earlier cooling events. For example, in northwest Iceland cooling after the Holocene Thermal Maximum is associated with a reduction in Myriophyllum alterniflorum pollen at Kagaðarh oll and Barðalaekjartj€ orn (Eddud ottir et al. 2015(Eddud ottir et al. , 2016. The additional impacts of the deposition of the V-1477 tephra layer probably exacerbated the impacts of the long-term cooling; it very likely damaged vegetation and reduced light penetration by increasing turbidity in the lake. Furthermore, the deposition of such a thick tephra layer would have separated the bed of the lake from the water column, preventing the remobilization of sediment and nutrients into the water column and thereby limiting productivity (Telford et al. 2004). The rapid decrease in productivity may indicate that tephra was a driver of change in an ecosystem that was already becoming increasingly vulnerable during the LIA. The evidence suggests that from the start of the LIA it was climate, rather than human activity, that was the dominant driver of aquatic ecosystem changes, even after AD 1477 when the terrestrial system underwent rapid change and soils were increasingly reworked into the lake. The contrast between the terrestrial and the aquatic systems highlights their differing sensitivities to environmental drivers, which has significant implications for understanding past environmental systems dynamics in Iceland.

Conclusions
The Kalmanstj€ orn record contributes significantly to the understanding of vegetation change after settlement in Iceland. It is the first pollen record generated from a lake sequence to show catchment-scale preservation of woodland until the late 15th century and indicates that the Norse settlers in this region maintained a larger stock of woodland resources than exist today until long after settlement (c. 600 years) and the onset of the LIA (c. 200 years). The terrestrial pollen data indicate a largely stable ecosystem with little change in either the woodland or minor taxa, and little soil erosion until AD 1477, despite evidence for grazing herbivores in the landscape. After the deposition of the V-1477 tephra, woodland rapidly declined, the local area became dominated by grazing-resistant rangeland and wetland taxa, and soil erosion increased, all probably reflecting a change in human land management in the area such as a substantial intensification of grazing.
The Kalmanstj€ orn record also provides evidence supporting the idea that settlement quickly led to a substantial change in the lake ecosystem. As at Helluvaðstj€ orn, there is a lack of evidence for soil erosion, which suggests an alternative mechanism must be responsible, such as increased rates of nutrient cycling by livestock. It appears that, at this site, the palaeolimnological record is more sensitive than terrestrial pollen to the first impacts of settlement, a conclusion that may also be applicable in areas with more cryptic settlement histories.
The study has generated a comprehensive data set aimed at investigating multiple different drivers of environmental change in Iceland, with the aim of better disentangling the causal mechanisms contributing to system development. The terrestrial system was dominated by land-use decisions from Landn am onwards, whether that was a decision to preserve or to utilize the woodland resources. There is little evidence for terrestrial vegetation change as a direct response to climatic change. Widespread soil erosion began after AD 1477, but it is unclear whether deforestation or harsher climates drove this change. The aquatic system was also driven into a more productive state by the introduction of new sources of nutrients after settlement, but during the LIA the lake system became increasingly dominated by climate. Understanding the interactions between anthropogenic and climatic drivers remains a complex issue in palaeoenvironmental research, and Iceland represents an ideal location in which to further investigate these relationships Table S1. Tephra geochemical data for analysed tephra layers in the Kalmanstj€ orn record. Table S2. Tephra geochemical data for the BCR2g and Lipari standards that were analysed during sample analyses.