An overview of Holocene climate reconstructions in northernmost Fennoscandia

An overview of climate reconstructions considering summer air temperatures and effective precipitation is provided for northernmost Fennoscandia. During the earliest part of the Holocene (11,700–10,000 cal. BP), temperatures rose rapidly and were followed by mild, wet and variable conditions. An early major warming peaked around 9500 cal. BP, although many records indicate that the main Holocene warming first occurred about c. 8000 cal. BP. The sub-regional pattern of climate change suggests a defining influence of the westerlies and the North Cape Current. Non-analog climatic conditions and lags in vegetation responses to climate change may explain some of the discrepancies seen in the early Holocene between proxies. In contrast to the perceivable variable onset of the main Holocene warm period, maximum temperatures are relatively consistent between the records, indicating temperatures 1.5±0.5°C above present. Precipitation was generally high from 10,000 cal. BP but decreased towards 8000 cal. BP when dry climatic conditions became predominant. After a stable period 8000–6000 cal. BP a gradual cooling was initiated, with a more abrupt period of change 4500–3800 cal. BP when the warm and dry climate of the mid-Holocene changed into the cool, wet and unstable climate of the late Holocene. Modern conditions were reached c. 2800 cal. BP. The Holocene Thermal Maximum may be defined several different ways: as temperatures distinctly above modern delimited to 9500–4000 cal. BP; as peak temperatures 9500–6000 cal. BP; and/or as climax vegetation in the period 8000–4000 cal. BP. Prior to 8000 cal. BP vegetation probably lagged behind the warming, whereas in the period 8000–4000 cal. BP an equilibrium between climate and vegetation was established.


Introduction
Reconstructions of climate and climate divergence are a prerequisite to understanding many aspects of human livelihood in the Holocene. Previous research has often focussed on the resulting environmental change with an emphasis on vegetation developments and terrestrial species. Increasingly, there has been an interest in individual climate events, in particular the 8.2 BP cold event and the consequences for human societies.
There has been less engagement with the relevance and impact of climate as a long-term factor. As part of the research project "Stone Age Demographics" on demographic patterns in northern Norway this report seeks to provide an overview of both the general trends as well as the divergence in the available climate data from northern Fennoscandia.
There has also been a keen interest in the paleo-oceanography of the North Atlantic and Barents Sea, providing well documented understanding of changes in sea surface temperatures and the strength of the major sea currents during the Holocene (e.g. Sarnthein et al. 2003;Andersen et al. 2004a;Eldevik et al. 2014).
Unfortunately, paleoclimatic proxy-data are notoriously noisy, and it is often difficult to discern the "true" climatic signal. There are numerous reasons for this, e.g. local variations, poor modern analogues, dating errors, vegetation response lags, statistical errors and poor correspondence between proxies and single climatic parameters.
Different climate proxies will show dissimilar ecological tipping points regarding one or several climate parameters. Hence, one and the same climatic event could potentially be signaled as several temporally separate climate "events". Some investigations have tried to mitigate such uncertainties by using a large number of records, from which statically mean climatic trends can be drawn (e.g. Seppä et al. 2009; and for a larger region Sejrup et al. 2016). Even though such investigations are valuable, they heavily depend on one or a few types of proxy-data and focus on a single climatic parameter, most commonly derived from pollen-based transfer-functions. So, despite the large number of records, results would still be vulnerable to any bias following that specific approach (e.g. Paus, 2013). In addition, the "noise" may actually contain meaningful information as it represents different aspects of the past climate (Huntley 2012). A more inclusive approach to the array of available proxies will thus provide a better assessment of past climate conditions.
In the present investigation, the most relevant sites/investigations and the full spectra of climate proxy data is provided, including qualitative and/or temporally limited data not suitable for numerical or statistical analysis.
The main objectives for the present study are to provide an overview of both the general trends as well as the divergence in the available climate data from northern Fennoscandia, with focus on the Holocene Thermal Maximum (HTM), including the build-up and subsequent decline in average temperatures. In addition, potential forcing factors and atmospheric circulation that are in agreement with the observed patterns will be discussed. The overview is also meant to increase the accessibility of the data of this key region for larger scheme investigations, not least for the benefit of disciplines such as ecology and archaeology as well as paleoclimatic studies targeting more specific questions.

Regional setting
The area under investigation is limited to Fennoscandia north of c. 68°N and extending to c. 71°N at the northernmost point of the European mainland (Fig. 1). The latitudinal extension is between c. 14 to 36°E. Despite the high latitude, the region experiences a mild climate because warm Atlantic water is brought north by ocean currents. The warm waters off the west coast in combination with predominantly westerly winds give rise to large climatic variation within the investigation area, this being enhanced by the Scandes Mountains which place much of the interior and eastern parts of the investigation area in a rain shadow. The western coastal area (as far as Magerøya and the west coast of the Porsanger peninsula) experiences a maritime climate with high precipitation (>1500 mm/yr at the outer coast), cool summers and mild winters. Along the coast east of Magerøya, the climate is drier (precipitation <750 mm/yr) and a few degrees colder during both the summer and winter. The interior experiences a more continental climate with little precipitation (<500 mm/yr) and strong seasonal contrasts (Dannevig & Harstveit, 2013). The main climatic variance can thus be described by two axes: a SW-NE axis along the coast with declining precipitation and to a lesser degree declining temperature towards the northeast; and a coast-inland axis with declining precipitation and increasing seasonal contrasts, especially colder winters, towards the interior. The climatic conditions are exemplified in Table 1. The standard climate reference period 1961-1990 is used to define the modern/present climate conditions. Table 1. Modern climatic conditions (met.no 1961Modern climatic conditions (met.no -1990 From the western part of the investigation area there are numerous pollen transferfunction based climate reconstructions available: Lake Tibetanus near Torneträsk (Hammarlund et al. 2002), Lake Gauptjern and Lake Jervtjern in the Dividalen region (Jensen & Vorren 2008), as well as Lake Barheivatn (Bjune et al. 2004), Lake Dalmutladdo (ibid.), Lake Toskaljavri (Seppä & Birks 2002) and Lake Tsoulbmajavri (Seppä & Birks 2001) in the Kilpisjävri region. Several sites on Andøya  and Jansvatnet near Hammerfest (Birks et al. 2012) reconstruct climate for the earliest part of the Holocene. In the eastern part of the investigation area, there are Lake Ifjord in the NE Norwegian part of the investigation area (Seppä et al. 2002a), and Lake KP-2 and Lake Yarnyshnoe-3 on the Kola Peninsula (Seppä et al. 2008). In addition to these individually published records, July temperatures have been reconstructed from 23 pollen records in Norway and northernmost Fennoscandia (Seppä et al. 2009), as well as Fennoscandia as a whole based on 59 records (Sejrup et al. 2016). Investigations using diatom-and/or chironomid-based climate transfer-functions are also more common in the western part of the investigation area, around Torneträsk (Bigler et al. 2002(Bigler et al. , 2003(Bigler et al. , 2006Larocque and Hall 2004) and Kilpisjärvi (Korhola et al. 2000;Seppä et al. 2002b).
A more direct way of inferring past summer temperatures than through transferfunctions are tree-line studies based on pollen, macrofossils and/or megafossils. In addition, indicator species, the general vegetation type, and several other types of proxies are useful in climate reconstructions, especially for determining the effective precipitation (precipitation minus evapotranspiration). As these studies vary in character and often contain qualitative information they will be described in more detail (see paragraphs below). Dendroclimatology is not considered here as the method is less suitable for reconstructing long-term climatic trends.
The marine records considered here are those most relevant for determining the strength of the Norwegian Atlantic Current and its eastern continuation as the North Cape Current.
The most important paleoclimatic investigations from northernmost Fennoscandia concerning the temporal constraints and amplitudes of the warmest Holocene periods are presented in Table 2. The driest period of the Holocene and wet shifts are provided in Table 3, and cold spells and/or shorter periods with abrupt cooling are presented in Table   4. Non transfer-function based climate reconstructions from the investigation area are discussed below.

Pollen and macrofossils used in tree-line studies
In the Lake Torneträsk area in NW Sweden (see Fig. 1), the sediments of a number of lakes have been analysed. By analysing lakes at different altitudes, Barnekow (2000) concluded that pine and birch were growing at maximum altitude c. 6300-4500 cal. BP. When corrected for land uplift of 20-40 m, this corresponds to a temperature of 1.5-2°C above present. This was also a dry period with a lowering of the lake level of 1-1.5 m (Lake Badsjön). After 4500 cal. BP, a more humid and cooler climate caused pine to retreat downslope.
In Dividalen (see Fig. 1), north of Torneträsk, pollen and macrofossils have been investigated at several sites (Jensen & Vorren, 2008 (Seppä 1996). The local presence of pine indicates that July temperatures were at least 1.2-1.5°C higher than present. Both the birch and pine tree lines started to retreat c. 5500 cal. BP. Aquatic macrofossils from northern Finland indicate a warm early-and mid-Holocene c. 11,000-6500 cal. BP, with mean July temperatures 2℃ above modern (Väliranta et. al. 2015; end of warm period as indicated by Shala et al. 2017).
Further east on the Kola Peninsula pollen and diatom evidence from Chuna Lake indicate a climatic optimum between 9000 and 5400/5000 cal. BP with warm and dry conditions, followed by a gradual cooling and increase in moisture (Solovieva & Jones 2002 (Snyder et al. 2000). A pine tree-line near the Barents coast would represent a warming of c. 2°C above today (Gervais et al. 2002). In northern Russia macrofossil evidence show that the forest advanced to near the current arctic coastline between 10,000 and 8000 cal. BP (9000-7000 BP) and retreated to its present position between 4500 and 3000 cal. BP (4000-3000 BP), although the northern tree-line on the Kola Peninsula was established later than more easterly regions . Mean July temperatures along the northern coastline of Russia may have been 2.5°C to 7.0°C warmer than today. Given the near absence macrofossils north of the modern tree-line dated to before 9500 and after 4000 cal. BP, these dates are here used to delimit the HTM.

Megafossils
Megafossils from near the shore of Lake Pompe (999 m asl, named Lake Njulla in other publications) indicate local presence of pine, birch and alder 9500-9000 cal. BP, and for birch and alder also 6200-5200 cal. BP (Kullman 1999). The highest position of the pine tree-line occurred during the early Holocene and was c. 500 m above present, which suggest that summers were c. 2.3°C warmer than today (corrected for land-uplift with -0.6°C). After recession of high-mountain glaciers in the Tärna and Kebnekaise mountains, NW Sweden, the exposed macro-and megafossils of wood were collected and dated (Kullman & Öberg, 2015). During the interval 9500-8500 cal. BP pine and birch grew 600-700 m higher than today, which corresponds to summer temperatures 3. showing a stable climate with relatively little interannual variability in the tree-rings (Eronen et al. 1999). This period was followed by a shift towards a cooler, more humid and variable climate, with most rapid changes around c. 4500 cal. BP and 2500-2000 cal. BP.

Other climate proxies
Coastal to inland mires in Troms, NW Norway, show a transition from geogenic to ombrogenous peat in the time span 5100-2500 cal. BP, with a concentration in the interval 3600-2900 cal. BP (Vorren 2001). Around 2900 cal. BP (2750 BP) there was a major change to cooler and wetter conditions. A peat humification record from the Lofoten region, NW Norway, indicates a relatively stable warm and dry climate until 4500 cal. BP (Vorren et al. 2012). Especially cold and wet conditions occurred in the period 3800-3000 cal. BP, followed by a highly variable climate until 1400 cal. BP. Lake sediment properties from the same area also indicate potential regional cooling events 10,900-10,200 and 9200-8000, a stable and productive period 7800-4800 cal. BP, and a distinct shift towards a wetter and more variable climate initiated at 4300 cal. BP (Balascio & Bradley 2012 (Hammarlund & Edwards 1998;Hammarlund et al. 2002). Varved sediments from Lake Sarsjön, northern Sweden, reveals a distinct increase in mineral deposition at 3710 ± 86 cal. BP, probably associated with a change towards a more unstable and maritime type of climate (Snowball et al. 1999).

Interpreting the climatic records
All palaeoclimatic reconstructions are associated with some degree of uncertainty. Most organisms respond to a multitude of climatic (and other) parameters but are often used as a proxy for a single climatic parameter, which may bias the result (see Huntley et al. 2012). The most common method for reconstructing terrestrial climate is the pollenbased transfer-function method (e.g. Seppä et al. 2004). This method reconstructs past temperatures, but it has its limitations, and criticism largely follows three lines: 1) The modern, human impacted vegetation used for training sets are poor analogues for past conditions; 2) Pollen dispersal-deposition and vegetation response to climate are too complicated to be accurately captured by a few site-independent parameters; and, 3) There might be a lag of several thousands of years between the onset of long-lasting climate ameliorations (as in the early Holocene) and the establishment of equilibrium between vegetation and climate (Paus, 2013). Transfer functions for aquatic taxa as chironomids and diatoms would theoretically be a more direct measurement of climate conditions, but they also suffer from poor modern analogues and autocorrelation with terrestrial vegetation affecting nutrient input and pH (see Bigler et al. 2002;Nyman et al. 2008).
Latitudinal or altitudinal tree-line studies using macro-or megafossils are good indicators of the HTM, as the tree-line is closely correlated to the mean July temperature. Even as such it is not without potential bias. Early Holocene mountain sites have experienced substantial isostatic uplift, which must be extrapolated from coastal areas where it can be determined. Another assumption is that the lapse rate has remained constant. In the investigation area, the modern difference in tree-line from the outer coast to the western part of the inland area is about 400 m (Moen 1999), and while birch (Betula pubescens) correlates with a July mean of 12℃ in oceanic western Norway it drops to 10℃ in eastern Norway (Odland 1996). Variance in oceanity is thus likely to have a major impact on the tree-line as well as the vegetation composition as a whole. The increased summer insolation during the early Holocene would also give a higher increase in growing degree days in mountain areas compared to the lowlands, possibly even more enhanced by a stronger massenerhebung effect. These potential biases in climate reconstructions highlight the importance of using several types of proxies in order to minimise the risk of misinterpretation.

Structuring Holocene climate change
In order to summarize and provide a better overview of the climate change during the Holocene it has here been structured into periods with relatively uniform climatic conditions or trends. The most common, general division of the Holocene climate is into three partitions: an early cool period with increasing temperatures, a warm middle phase with maximum temperatures, and a cool late phase with declining temperatures (e.g. von Post 1946;Firbas 1949;Hafsten 1970;Wanner et al. 2011). On a general level the records considered here fit well within this scheme, but at closer examination problems arise as different climate characteristics are out of phase. Because of this a more neutral labelling of the Holocene into an early, middle and late period is applied. Each of these periods are then divided into an early (EP) and late phase (LP) based on the climatic characteristics.
With the exception of the start of the Holocene and the well dated cold spell at 2800 cal.
BP the climatic zones are set to even millennia. The climatic conditions during the different periods are described below and have been summarised in Table 5, and the schematic temperature development is presented in Fig. 2.
Early Holocene EP, 11,700-10,000 cal. BP, cold, rapid warming, -2 to +1°C At the onset of the Holocene ca. 11,700 cal. BP the Scandinavian Ice Sheet covered most of the inner fjord area and all of the inland (Romundset et al. 2017). At the end of the period it had dwindled down to a fraction of its former size situated SSW of the investigation area (Stroeven et al. 2016). June insolation would have been at its peak 12,000-11,000 cal. BP, and July insolation 10,000-9000 cal. BP (65°N, Berger & Loutre 1991;Berger 1992). Vegetation development during the early Holocene was strongly delayed by retreating glaciers, katabatic winds, soil development and migration, showing strong local and regional variance. At the coast, this delay is especially evident prior to c.
11,000 cal. BP, while this situation continues longer in the inland closer to the ice margin.
The dry conditions of the Younger Dryas prevailed some time into the early Holocene (Birks 2015 -2°C) at the start of the Holocene (11,700 cal. BP) to temperatures at or slightly above present (+1°C) around 10,000 cal. BP. However, the climate is variable and there are several severe setbacks in the warming recorded (e.g. Seppä et al. 2002a). Notably, aquatic macrofossils in general indicate about 2°C higher temperatures than pollen-based studies in the region, and record temperatures 2°C above present already at 11,000 cal. BP (Väliranta et al. 2015). It is possible that terrestrial vegetation was not in equilibrium with the climate because of soil maturation and migrational lag, or that the microclimate of the aquatic taxa rather reflects the high solar insolation than air temperatures and/or a shorter but warmer growing season. Climatic reconstructions for this period are relatively sparse, and the environment is heavily influenced by the rapidly melting Scandinavian Ice Sheet. Nevertheless, the general picture is that of a variable climate with rapidly increasing temperature and a shift from the dry Younger Dryas conditions towards increased precipitation.
Early Holocene LP, 10,000-8000 cal. BP, wet, warm and variable, +0 to +2°C Tree-birch started to spread c. 11,000 cal. BP, and around 10,000 cal. BP most of the early Holocene shrub-tundra had been replaced by oceanic birch-forests across northernmost Fennoscandia, indicating a shift towards a relatively mild and moist climate strongly influenced by westerlies (Seppä & Hammarlund 2000;Sjögren & Damm 2018). Megafossils from pine indicate that the Holocene pine forest-line peaked in altitude during the period 9500-9000 cal. BP, corresponding to summer temperatures 2-3°C above present (Kullman 1999;Kullman & Öberg 2015). An increase in sea surface temperature is evident in most marine records ranging from 10,700 to 9000 cal. BP, and a start of the HTM around 10,000 cal. BP can also be seen in records from Greenland and northern Russia. A climatic shift seems to occur around 9000-8500 cal. BP, when the megafossils indicate a decrease in summer temperatures, and other investigations suggest that the climate became drier (Eronen et al. 1999;Hyvärinen & Alhonen 1994;Allen et al. 2007), most likely including a reduction in winter precipitation (Hammarlund & Edwards 1998;Bakke et al. 2005). A few other records indicate a cooling around 9000 cal. BP (Korhola et al. 2000;Sarnthein et al. 2003;Mangerud & Svendsen 2018). On the other hand, several records also indicate a warming at the same time (Bigler et al. 2002;Solovieva & Jones 2002;Allen et al. 2007).
Notably, even though pine abundance did not reach its maximum (as determined by pollen and macrofossils) until later, it did expand in the period after 10,000 cal. BP and especially after 9000 cal. BP (Seppä & Hammarlund 2000;Snyder et al. 2000;Gervais et al. 2002;Sjögren & Damm 2018). Mega-and macrofossils of pine dated to c. 9500 cal. BP are spread across the region indicating a peak occurrence at this time (Kullman 1999, Kullman & Öberg 2015, Väliranta et al. 2015Virdnejávri Ua-46463, Marianne Skandfer, personal communication 2017). Overall, the period is characterized by warm, variable and wet conditions as indicated by many records, with peak temperatures around 9500 cal. BP.
Although, not all records agree with this, suggesting a complex environmental context at the time which prevents firm conclusions.

Mid Holocene EP, 8000-6000 cal. BP, dry, warm and stable, +1 to 2°C
Most transfer-function based temperature reconstructions indicate that the period with highest summer temperatures started around 8000 cal. BP. Pine (Pinus sylvestris) and alder (Alnus incana) became common and replaced much of the previous birch-fern forest (Sjögren & Damm 2018), and the tree-line reached its highest altitude and latitude in some areas. Many records also indicate drier conditions and there are few indications of climatic fluctuations in the records. The overall impression of this period is that it was warm, dry and stable, and the warmest and climatically most stable millennia of the Holocene occurred 8000-7000 cal. BP (e.g. Seppä et al. 2009).
Mid Holocene LP, 6000-4000 cal. BP, dry and warm, +0.5 to 1.5°C The general climate is similar to the proceeding period, though deviating by showing records of increasingly colder and more unstable conditions. A major cold event seems to occur around 5500 cal. BP (see Table 4). Some records based on the extension of pine indicate the highest temperature in this period (Barnekow 2000;Helama et al. 2004), but the peak abundance of pine could be explained by reduced precipitation, local variations and/or migratory lags, rather than temperature alone. The shift between the warm mid Holocene and cold late Holocene could not be pinpointed down to a single century but to a transition phase 4500-3800 cal. BP, even though the dates 4500, 4200 and 3800 recur in many investigations. Here the even time point 4000 cal. BP is used as a simplified definition for this gradual or step-wise change. (Berger & Loutre 1991) and the late Holocene increase in effective precipitation as exemplified by the Rystad 1 humification record (Vorren et al. 2012). The schematic temperature curve is the general deviation of July air temperature compared to modern as deduced from the available climate records (see text and Table 2). Dashed line indicates a more uncertain development with more spatial variation, larger discrepancies between proxies and fewer continuous reconstructions. The grey fields indicate the common range within and between records. Cold spells / rapid cooling declines as reflected in the terrestrial climatic records (see Table 4).

Late Holocene EP, 4000-2800 cal. BP, wet and variable +0 to 1°C
The long-term gradual decline in summer temperatures became more abrupt around 4000 cal. BP, but the change towards much wetter conditions 4500-4000 cal. BP was more profound (see Table 3). Most likely the gradually declining solar insolation and the associated reforming of Arctic sea-ice reached a threshold that resulted in a major reconfiguration of the climatic system (see e.g. Coumou et al. 2018).
Late Holocene LP, 2800-0 cal. BP, wet, cool and variable, -0.5 to +0.5°C In competition with the 8.2 event, the most pronounced cold spell occurred c. 2800 cal.
BP, marking the end of the warmer climate and the start of the modern climatic conditions. Notably though, many proxy records show no indications of these cold spells which suggests that the abrupt decline in temperature was not too severe and/or of short 1400 cal. BP (see Helama et al. 2017) and the Little Ice Age c. 350-100 cal. BP (Lamb 1995).

The onset of the Holocene Thermal Maximum
With few exceptions, the peak temperature of the Holocene is fairly similar between proxies, about 1.5±0.5℃ higher than present. This is in line with estimates of HTM temperatures in northern North America (north of 60°N) that on average measure 1.6±0.8℃ above present (Kaufman et al. 2004). In the climate reconstructions considered here, some discrepancies between proxies become obvious. In Lake Yarnyshnoe-3 on the Kola Peninsula pine stomata (Snyder et al. 2000) indicate the presence of pine some 100 km north of its current distribution between c. 9000 and 4500 cal. BP, representing a summer temperature 2°C above present. In Poteryanny Zub, another lake on the Kola Peninsula, stomata indicate local presence of pine c. 9000 although no expansion is seen in the pollen record until 8000 cal. BP (Gervais et al. 2002). In contrast, the pollen-based transfer-function temperature reconstructions from these two lakes indicate a HTM around 7500-6500 cal. BP with temperature 1.5°C above present (Seppä et al., 2008). In the western part of the investigation area, pine was growing at high altitude in the Scandes from about 9500 cal. BP (Kullman 1999;Kullman & Öberg 2015), while most pollen-based transfer-function temperature reconstructions indicate maximum temperatures no earlier than 8000 cal. BP (see Table 2). So, a lag of c.
climax vegetation / pollen-based transfer-function temperature reconstructions. This lag of 1500 years is of similar magnitude as observed in mountain areas in mid-Scandinavia (Paus, 2013). Hence, tree-line studies across the region indicate a warming of c. +2°C some 1500 years prior to the bulk of pollen-based transfer-function temperature reconstructions.
Non-analogue climate (e.g. Huntley 2012) and soil maturation (e.g. Pennington 1986) may explain some of these discrepancies. Pollen-based transfer-function estimates for both terrestrial and aquatic taxa have presented a poor fit for the early Holocene temperatures (e.g. Bigler et al. 2002;Nyman et al. 2008, Seppä et al. 2008. Model-based quantitative vegetation reconstructions from the Dividalen area have showed that changes in July temperature could explain changes in the vegetation after 7400 cal. BP, but not prior to 7400 cal. BP, suggesting that other factors were decisive for the vegetation dynamic in the early Holocene (Sjögren et al. 2015). The main reason for these discrepancies is likely the non-analogue climate conditions of the early Holocene, particularly concerning solar insolation. Peak summer solar insolation at 70°N occurred c. 11,000 cal. BP and peak annual insolation 11,000-8000 cal. BP (Huybers 2006;Huybers & Eisenman 2006). At 11,000 cal. BP, annual (integrated) insolation was only 2% above modern. Insolation above 275 W/m 2 , roughly equivalent with the insolation that today is required to reach temperatures above 0°C, was 5% higher. Although peak summer insolation, today between 475-500 W/m 2 and equivalent to a monthly mean temperature around 10°C at the northern coast, was 82% higher. Combined with a strong influence of the westerlies, this would have resulted in non-analogue climate conditions with warm and wet summers combined with long, cold and snow-rich winters. Possibly, the tree-line would have responded more positively to these conditions than the forest-line/density, partially explaining the discrepancies between proxies. Nevertheless, climate change will inevitably occur prior to vegetation change, resulting in a shorter or longer lag. When different proxies / records show asynchronous responses to climate change, it should be the earliest records that provide the most likely date for the onset of change (allowing for dating error and comparable quality), in this case 9500 cal. BP.

The Holocene Thermal Decline
From about 7000-6000 cal. BP there is a general decrease in July temperature which has been attributed to the gradual decline in solar insolation (Sejrup et al. 2016, Zhang et al. 2017, although the period 6000-5000 also coincided with low solar activity (Mayewski et al. 2004). Colder climate and contrasting patterns of hydrological change have been recognised worldwide in the period 5600-5000 cal. BP (Magny & Haas 2004), including at Spitsbergen c. 5500 cal. BP (Alsos et al. 2016). A striking example from Scandinavia is the sudden extinction of the pond turtle (Emys orbicularis) from southern Sweden around  Table 3 and 4). The cold and wet conditions of the late Holocene are also associated with a more variable climate (e.g. Linge et al. 2009;Balascio and Bradley 2012;Vorren et al. 2012). The high-resolution record from Lake Igelsjön further south in Sweden reveals that this transition occurred in two steps: the first between 4450 and 4350 cal. BP and the second and more pronounced between 4000 and 3800 cal. BP (Jessen et al. 2005

Spatial variation
During the early Holocene, the expansion of pine was time transgressive, ranging from 10,900-8900 cal. BP in the north-east to 8800-6300 cal. BP in the west and south-west (Seppä & Hammarlund, 2000). Sjögren & Damm (2018)  A key site to unravel the climate conditions is Lake Tibetanus in the Torneträsk area, NW Sweden, with a sediment record going back 10,000 yr (Berglund et al. 1996;Hammarlund & Edwards 1998;Hammarlund et al. 2002). The oxygen isotope and pollen records indicate moist, maritime conditions during the early part of the record. Maximum influence of the westerlies (high zonal index) occurred c. 9500 cal. BP, then the influence rapidly declined until 8800 cal. BP, and more gradually until 6300 cal. BP. The main vegetation changes closely follow the decline in precipitation, and pollen inferred temperature is inverse to precipitation over the early part of the record, prior to 6500 cal. BP. The driest conditions of the Holocene seem to have occurred c. 6500-6000 cal. BP (Hyvärinen & Alhonen 1994;Hammarlund et al. 2002), when the Norwegian Atlantic Current was strongly weakened (Andersen et al. 2004a). The reduced northward flow of warm surface water would by itself indicate a weakening of the westerlies, and colder water would also reduce the moisture uptake, suggesting (relatively) weaker and drier westerlies. This is concurrent with the expansion of pine in some areas (Barnekow 2000;Helama et al. 2004), the last major expansion of pine during the Holocene.
The early presence of pine in high altitude areas of the Scandes (Kullman 1999;Kullman & Öberg 2015) seemingly contradicts a peak influence of western air masses at the time (Hammarlund & Edwards 1998). One possible explanation is topography, i.e. that pine grows in areas largely sheltered from the westerlies, and the finds normally occur in the climatically most favourable areas (Kullman, 2013). A second explanation is that the strength of the westerlies was very variable. Interestingly, Huntley et al. (2013) report coincidental peaks in dwarf birch pollen (Betula nana) in two lakes c. 9600-9500 and 8900 cal. BP, which they interpret as cooling events caused by a weakening of the North Cape Current. As this current is largely driven by the strength of the westerlies it would indicate a reduction of western air masses and a more continental climate, favouring the establishment of pine.
The North Cape Current would have a direct influence on the NE-E coastal part of the investigation area. Interestingly, the warmest condition of the eastern Barents Sea (7800-6800 cal. BP, Duplessy et al. 2001Duplessy et al. , 2005 coincides with the pollen-based transfer-function derived temperature peak at the north coast of the Kola Peninsula (7500-6500 cal. BP., Seppä et al. 2008). The relatively late peak in temperatures of the eastern Barents Sea has been attributed to the melting of local glaciers (Duplessy et al. 2001(Duplessy et al. , 2005, and the cold coastal water of the Kola coast might have had a restraining effect on the establishment of pine forests compared to NE Norway. The sub-regional climate variation in northernmost Fennoscandia can thus largely be explained by spatial and temporal variation in the influence of westerly air masses.

Forcing factors
The climate dynamic of the Holocene could largely be explained by two forcing factorssummer solar insolation and polar ice. In the earliest part of the Holocene the climate was strongly influenced by the remaining ice sheets. Meltwater bursts, especially in the earliest period prior to 10,000 cal. BP, would disrupt the sea currents causing temporary cold spells and a more Arctic climate. Even though summer solar insolation was high, the remaining large ice sheets in the Arctic would maintain a high latitudinal temperature gradient that would fuel a zonal atmospheric circulation with strong westerlies. After most of the inland ice sheets were gone this would result in a maritime climate, but due to the high summer solar insolation the summers would nonetheless be warm. During the middle Holocene summer solar insolation was still rather high, and with the inland icesheets completely gone and with presumably little sea-ice in the Arctic the altitudinal temperature gradient would be low. This would result in weak westerlies and a more meridional atmospheric circulation, i.e. drier and more continental conditions (see Yu and Harrison, 1995). During the late Holocene the weak summer solar insolation together with the reforming of glaciers and especially the expansion of Arctic sea-ice would yet again increase the latitudinal temperature gradient. The result would be a more zonal atmospheric circulation with increased influence of the westerlies. This would again give rise to a mild and wet climate, but with the low summer insolation of the late Holocene summers would be cool. If this is used as an analogue for future climate change, the current loss of Arctic sea-ice caused by global warming would weaken the latitudinal gradient, which could result in a new, major reconfiguration of the atmospheric circulation to a more meridional type (see Coumou et al. 2018).

Acknowledgement
Charlotte Damm and Aage Paus have made insightful comments and valuable contributions to the report. Anonymous reviews have improved earlier manuscript versions. Roberta Gordaoff contributed with language revision. The report is a contribution to the Stone Age Demographics project at UiT -The Arctic University of Norway (NFR grant number 261760).