Holocene glacial history of Svalbard: Status, perspectives and challenges

We synthesize the current understanding of glacier activity on Svalbard from the end of the Late Pleistocene (12,000 yrs. before present) to the end of the Little Ice Age ( c. 1920 AD). Our glacier history is derived from the SVALHOLA database, the ﬁ rst compilation of Holocene geochronology for Svalbard and the surrounding waters, including over 1,800 radiocarbon, terrestrial cosmogenic nuclide and optically stimulated luminescence ages. Data have been categorized by geological setting, uniformly (re-)calibrated, quality assessed and ultimately used to constrain glacier ﬂ uctuations (deglaciation, ice free conditions, glacier re-advances and ice marginal positions). We advance existing knowledge by mapping the extent and distribution of ice-cover during the Holocene glacial maximum and the glacial minimum, as well as present retreat rates (and percentages) within Early Holocene fjord-systems. Throughout the Holocene, Svalbard glaciers have responded to a varying combination of climatic, environmental and dynamic driving factors which in ﬂ uence both the extent and behavior of ice margins. We discuss the complexities of glacier systems and their dynamics in response to changes in climate. This review provides a holistic state of the art of Holocene glaciers on Svalbard, suitable for orienting future works which address gaps in our current knowledge.


Background
Syntheses of accumulated field and geochronology data are prerequisite for placing new observations in context and for re-assessing existing interpretations (e.g., Hughes et al., 2016).Empirical constraints are critical for developing and calibrating models that simulate palaeoclimate as well as past glacier dynamics and processes (Patton et al., 2017).Within this context, we synthesize published literature of Holocene glacier history from the Svalbard archipelago and its surrounding waters.We introduce a database of empirical data specific to the Svalbard region as well as summarize prominent glacier fluctuations through the last 12 ka BP (kiloyears before present; Fig. 1).
The North Atlantic Holocene climate varied considerably, with fluctuations between warm and cold as well as humid and dry conditions on the multidecadal to multicentennial timescale having a direct effect on glacier-systems (Mayewski et al., 2004;Wanner et al., 2011).'Arctic amplification' refers to the sensitivity of the Arctic to these climatic fluctuations and, specifically, how the Arctic is significantly affected by minor shifts in temperature and precipitation compared to lower latitudes (Masso-Delmotte et al. 2013).By studying the rate and  (Jakobsson et al., 2012) and TopoSvalbard © Norwegian Polar Institute 2020 respectively.magnitude of past cryosphere changes we are able to better understand current dynamics, as well as more effectively predict antecedent climate scenarios (McKay and Kaufman, 2014).
We reconstruct Svalbard glacier history by synthesizing geochronological data presented in scientific publications detailing Holocene sedimentary archives (marine, terrestrial and lacustrine) and landforms.Our synthesis leads us up to the earliest historical ice front observations from Svalbard, and places this accumulated Holocene data and observations in context with the ever-developing understanding of ice dynamics and the Arctic climate system.We strive to not only distinguish spatial trends in ice expansion and decay of Svalbard glaciers and ice caps through the Holocene.Furthermore, we discuss how reconstructed phases of ice-marginal change relate to palaeo-environmental conditions, e.g.climate, relative sea level, internal glacial processes or some combination.Understanding Arctic Holocene history can provide insight as to how future changes may influence glaciers, sea level and environmental conditions.
Svalbard has a long and rich history of scientific observations.The first field observations of Svalbard coincide with the earliest undisputed discovery of "Spitsbergen" by Willem Barentsz in 1596 (Claesz, 1598;Hacquebord, 1995;Arlov, 2005).Although early settlers were initially drawn north to the coasts of Svalbard for hunting, whaling and trapping, in the first half of the 1800s by the mid-1800s scientific expeditions began accumulating detailed observations of the landscape, climate and ice-cover around the archipelago (Lottin et al., 1842;Bertrand, 1852).Through the second half of 19th century, numerous international scientific expeditions visited Svalbard eventually culminating in the first International Polar Year in 1882-1883 (Ekholm, 1887; Fig. 2).During the last c.150 years, the Svalbard archipelago has become a natural laboratory for observing glacial landforms, processes and dynamics as well as the terrain which ice masses have shaped through the Quaternary (Holmström, 1865;Ingólfsson, 2011;Ingólfsson and Landvik, 2013).The bulk of this work has focused on constraining the Weichselian growth and break-up of the marine-based Svalbard-Barents Sea Ice Sheet (SBSIS; Blake, 1962;Boulton, 1979;Österholm, 1990;Landvik et al., 1998;Mangerud et al., 1998;Gjermundsen et al., 2013;Hormes et al., 2013;Ingólfsson and Landvik, 2013;Hughes et al., 2016).These reviews predominantly focus on Last Glacial Maximum (LGM) ice configuration, timing and disintegration, often excluding Holocene data.However, several studies of postglacial sea level and subsequent isostatic uplift have been used to better understand LGM ice cover and previous centers of mass.These works have indirectly summarized chronological shoreline development through the Holocene (e.g.Salvigsen, 1981;Forman, 1990;Forman et al., 2004;Bondevik et al., 1995;Forman et al., 2004;Sessford et al., 2015;Schomacker et al., 2019).
Holocene studies have long targeted unknowns of the glacier and climate history of Svalbard by mapping (terrestrial and marine) glacier landforms and deposits as well as investigations of lacustrine sediments from threshold lakes.Our synthesis addresses the following outstanding research questions: (1) When were Svalbard glaciers at their Holocene maximum and minimum ice extent and how extensive were they during these periods?(2) When were Svalbard's glaciers retreating / advancing, and what were the key drivers controlling Holocene glacier activity?(3) To what extent has ice dynamics and surge-type behavior influenced Holocene glacier fluctuations on Svalbard?

Setting and climate
Located along the dominant corridor of atmospheric moisture, between the Atlantic and the Arctic Basin, Svalbard spans from 74°to 81°N (Fig. 1; Drange et al., 2005).At present, glaciers and ice caps cover roughly 57 % of the archipelago (Nuth et al., 2013).The region has a sensitive climate due to its position at the northern extent of the North Atlantic Drift (West Spitsbergen Current; Fig. 1A) and the southern border of the Arctic sea-ice front (Aagaard et al., 1987;Rogers et al., 2005).Svalbard is categorized as having a dry High Arctic climate with periglacial conditions, extreme winter temperatures and continuous permafrost (French, 2007;Christiansen et al., 2010).
Despite its high latitude, Svalbard currently experiences a relatively mild climate where the warm West Spitsbergen Current flows along the western coast of Spitsbergen and influences weather and sea ice conditions (Fig. 1; Førland et al., 1997).Regional climate is influenced by the interactions between the Icelandic Low and Siberian High pressure systems and as a result high temperatures are driven north over Svalbard by the main North Atlantic cyclone track (Hanssen-Bauer et al., 1990).Svalbard precipitation is closely coupled to the mode of the North Atlantic Oscillation (NAO; Dickson et al., 2000) and falls predominantly in solid form (Eckerstorfer and Christiansen, 2011).Today, the interactions of these air masses along the western margin of the archipelago commonly drive winter conditions with warmer and wetter climate than normally expected at such latitudes (Førland et al., 1997;Eckerstorfer and Christiansen, 2011).

Sedimentary archives, landforms and geological reconstructions
Holocene reconstructions of glaciers represent a mosaic of data developed through a suite of sedimentary archives and landforms present in both marine and terrestrial environments.The different stratigraphic archives used to reconstruct glacier and climate history include terrestrial geological sections, marine sediment cores and thresholdlake sediment records (Fig. 3).Geophysical data such as ground penetrating radar or sub-bottom profiler records can be paired with sediment cores and terrestrial stratigraphy to extrapolate across larger areas.Additionally, the relative (and absolute) age of landforms and their cross-cutting relationships identified in submarine and subaerial data (marine bathymetry and aerial imagery) are used to reconstruct past glacier extent, dynamics and other environmental conditions like relative sea level (Fig. 3).Moraines formed on land or at the margin of a tidewater glacier have been identified and correspond to glacier readvances (or still-stands).Raised marine sediments fingerprint postglacial coastlines around Svalbard, in some cases containing datable drift material (i.e.mollusc shells, whalebones and driftwood).Ice-core stratigraphy is discussed based on 10 studies from 12 ice cores taken on Svalbard, however it is not the focus of this review given the relatively short and young cryostratigraphic record (less than 1 ka BP; Isaksson et al., 2003Isaksson et al., , 2005;;Grinsted et al., 2009;Divine et al., 2011).

Compilation of ages
Age determinations from Svalbard and the surrounding waters that provide chronological constraints on glacier cover and marginal fluctuations or insight into climatic conditions over the last 12.0 ka were compiled into the SVALHOLA database (Table S1).Compiled ages and their meta-data were extracted from scientific publications, books, doctoral theses, geological reports, and maps (Table 1).Dates obtained from compilations are cited as well as the original source.To assure quality and consistency, ages from M.Sc.theses were excluded unless subsequently referenced in a peer-reviewed publication.The SVALH-OLA database compiles previously published ages of radiocarbon ( 14 C, 14 C AMS, 14 C Conv.), luminescence (optically and infrared stimulated, OSL and IRSL) and terrestrial cosmogenic nuclide (TCN 10 Be, TCN 26 Al, TCN 36 Cl).Thermoluminescence is nowadays considered less suitable for sediment dating and is not included in the database (Fuchs and Owen, 2008;Wintle, 2008).The database omits 210 Pb chronologies as the review does not focuses the post-LIA period.
The database has a census date of the 1 st May, 2020 (which will be updated to the date of publication).All ages presented in the text are referred to in kiloyears (ka) unless otherwise stated.All radiocarbon ages have been calibrated and are referred to in ka before present (ka BP; present = 1950 AD), whereas the luminescence and cosmogenic nuclide ages are referred to in ka.

Radiocarbon ages
To present and evaluate the SVALHOLA dataset consistently, all radiocarbon dates were recalibrated with INTCAL13 and MARINE13 calibration curves using the Calib Rev. 7.0.4program (Reimer et al., 2013).Calibrated age ranges are presented as both 1σ (68% probability range) and 2σ (95% probability range) in the database.We use the centre of the age distribution ± half of the total 2σ range to represent the calibrated uncertainty for each age in our review.Given the past variations in correction and calibration for the geochronological methods, we extend the database to include ages with calibrated 2σ ranges where the upper 2σ falls above 12.0 cal.ka BP, but the full 2σ range indicating that the age potentially is younger than 12.0 ka BP.The ages presented in text and figures are calibrated center ages in kiloyears before present (ka BP) unless otherwise clarified.
A marine reservoir age has been implemented for all marine samples by selecting 'MARINE13' and using a regional "Delta R" (ΔR) value of 70 ± 30 (Table 1, S1; Mangerud et al., 2006;Mangerud and Svendsen, 2017).Our chosen ΔR of 70 ± 30 is specific for the Svalbard region (Mangerud et al., 2006;Mangerud and Svendsen, 2017) and deviates from the DATED1 compilation, which for simplicity utilized a ΔR value of 0 for all marine samples from their reconstruction of the last Eurasian ice sheets (Hughes et al., 2016).Both radiocarbon ages and calibrated ages are presented in the SVALHOLA database (Table S1).

Terrestrial cosmogenic exposure ages
A cosmogenic exposure age represents the cumulative time the sampled material has been exposed to cosmic radiation and need not always correspond to an actual age of an event.However, for the dates included in the SVALHOLA databasesamples are almost exclusively erratic boulder ages that target moraines and relate to an ice marginal position.Within the Holocene, we assume that an exposure age represents one continuous exposure history.
As argued by Hughes et al. (2016) and Young et al. (2018), no corrections have been made for post exposure uplift, erosion, or snow cover.Cosmogenic ages are given in ka prior to the year of sampling.We have chosen not to recalculate the ages to ka before present (BP) and thus there may be roughly a 50-year offset from the radiocarbon ages.

Luminescence ages
Luminescence dating determines the last time a quartz or feldspar grain was exposed to sunlight, which is usually taken to be the time of deposition of the sampled sediment (Wintle, 2008).A luminescence age is given in ka prior to the year of sampling (Alexanderson and Murray, 2012a).Additionally, we have chosen not to recalculated the ages to ka before present (BP) and thus there may be roughly a 50-year offset from the radiocarbon ages.In the SVALHOLA database, ages are listed as presented in original publications (Table S2).For quality assessment criteria of OSL and IRSL ages, see Table 2.

Consistency and quality assessment of dates
The SVALHOLA database consists of accumulated geochronological information acquired over the last 50 years.Dating techniques have developed over this period and the standard of what is considered "reliable" has risen (Ivy-Ochs and Briner, 2014;Hughes et al., 2016).We utilize the structure of the quality assessment criteria introduced in the DATED1 database (Table 2) to characterize the reliability of each age in the SVALHOLA database (Hughes et al., 2016).We have rated all ages depending on the dating technique on a 3-point system (quality 1-3) to rule-out potentially misleading ages.According to Hughes et al. (2016) a quality mark of; 1 = all criteria is met (likely reliable), 2 = some of the criteria are met but not all (probably reliable), 3 = no

Table 1
Presented metadata recorded for each date included in the database (Table S1).Metadata from the database form the core of the criteria for quality assessment and palaeoglaciological classifications of each date as defined by the DATED1 database (Modified from Hughes et al., 2016).

Quality control
• Reliability of the age: 1 = likely reliable; 2 = probably reliable; 3 = likely unreliable (see Table 2 for criteria)

Ages
• Uncalibrated radiocarbon age / error (as reported, without marine correction or reservoir effect) • TCN age and error (as reported in source) • OSL age and error (as reported in source) • Calibrated/calendar age and error (reported to 2σ).Radiocarbon ages calibrated to INTCAL13 or MARINE13 (Reimer et al., 2013) as appropriate (on basis of type of organic material: T/M). 10 Be and 26 Al TCN ages recalculated using 'Arctic' production rate (Young et al., 2013) and Lal/Stone scaling (Lal, 1991;Stone, 2000).Necessary information to recalculate 10 Be and 26 Al TCN ages using different production rates additionally collated and recorded in Table S3 • Age label; median age is rounded to nearest 10 th as refered to in text.
• Comments on calibration (e.g.beyond calibration curve limit)

Comments
• Any additional pertinent comments (e.g.reliability of date) Citation information • Source reference (author, year) • Compilation reference (author, year) • Database reference (e.g., Hormes et al., 2013; Mangerud and Svendsen, 2017)  criteria are met (likely unreliable).Dates suggested to be likely unreliable by original authors or other subsequent databases or compilations are rated quality 3. Dates of quality 1 and 2 standard have been used to reconstruct glacial history while ages deemed quality 3 are presented in Table S1, but have been excluded from the developed reconstruction.We deviate from the DATED1 standard in one key aspect; radiocarbon ages from sediment-feeding marine molluscs (i.e.Portlandia arctica) are considered likely unreliable.We strictly rate them as quality 3, thus excluding them from the reconstruction due to the high probability of an exaggerated age (England et al., 2013;Hughes et al., 2016).

Ice free
We have simplified the palaeoglaciological/climatological significance for a portion of the ages compiled in the SVALHOLA database founded upon metadata from the publications or the author's suggestions presented in the different studies.We assume all ages indicate icefree or non-glaciated conditions unless otherwise stated.In some cases, additional classifications are placed on the ages.We do not classify the palaeoglaciological/climatological significance of individual ages taken from marine or lacustrine cores, but address their chronologies as a whole in both the results and discussion.

Re-advance
Ages of dateable material (e.g.vegetation, shell or bone with 14 C or sand with luminescence) that has been reworked or overridden by an advancing glacier are classified as maximum constraining ages for a glacier re-advance (Baranowski and Karlén, 1976;Ronnert and Landvik, 1993;Kristensen et al., 2009).In a study where numerous samples have been dated to constrain one glacier re-advance, all ages are classified as a "re-advance" in the SVALHOLA database, but only the best constraining age (i.e. the youngest) is presented in the figures (Humlum et al., 2005;Table S1).While older ages may inform about when the glacier was smaller than present (Punning et al., 1976;Kristensen et al., 2009;Lyså et al., 2018;Larsen et al., 2018), each re-advance age presented in figures reflects a specific event for a specific glacier margin.Few Holocene glacier advances have been constrained in time by both maximum and minimum limiting ages (Lønne, 2005;Farnsworth et al., 2018).Based on consideration by original authors, ages classified as a "re-advance" are interpreted to correspond to a glacier fluctuation at or shortly following the mid-point of the youngest age (within the error margin).

Deglacial
We classify ages as "deglacial" where the site possesses stratigraphic information suggesting ice-free conditions shortly following cover by an ice sheet or a glacier.Such dates include basal organic material from lake cores, the lowermost shells or foraminifera in marine cores from fjords and raised glacial marine sediments from fjord-heads (Landvik et al., 1987;Mangerud et al., 1992;Svendsen and Mangerud, 1997;Hald et al., 2004;Alsos et al., 2015;Bartels et al., 2018;Farnsworth et al., 2018;Larsen et al., 2018).Additionally, the classification is used for radiocarbon ages of dateable material that has been sampled up-ice of modern ice margins (Blake, 1989;Oerlemans et al., 2011) suggesting reduced ice-cover.Unlike previous reviews (Landvik et al., 1998;Ingólfsson and Landvik, 2013;Hormes et al., 2013;Hughes et al., 2016;Hogan et al., 2017) we choose not to focus on the deglaciation in this paper but rather ice extent during the transition from Late Pleistocene to Early Holocene.

Margin
An age is classified as "margin" if the dated material is associated with an ice-marginal position.In the SVALHOLA database this is mainly erratic boulder ages, but could also be radiocarbon or luminescence ages relating to e.g. an ice-contact delta.Unlike ages associated with a re-advance, where the youngest age is considered the best constraint, exposure ages corresponding to an ice margin are presented in cumulative probability or in a histogram.

Environmental associations
We assign associated environmental conditions to two types of samples: thermophilous marine molluscs and terrestrial plants that

Dating technique
Quality control criteria Radiocarbon 14 C Conv (Conventional), 14 C AMS • Known and uncontaminated sample material; sediment-feeding marine mollusc (e.g.Portlandia arctica) receives the lowest rating • Organic content > 5% LOI • Sample composition: Conv -bulk samples not acceptable; AMS -bulk sample acceptable • Within calibration range of INTCAL13/MARINE13 • Uncalibrated 14 C age determination provided with errors to enable recalibration using the latest calibration curves • Multiple and/or stratigraphically consistent ages Terrestrial cosmogenic nuclide TCN 10 Be, 26 Al, 36 Cl • Multiple (ideally three or more, but at least two) samples from the same feature/site • Ages are internally consistent and clustered (reduced Chi-square value ~1) • Observed spread in ages is similar to expected measurement uncertainty • Geomorphological setting is accounted for: erosion, submergence, uplift • Data necessary to recalculate ages ( 10 Be, 26 Al) using different production rates (Balco et al. 2008)  • No indication of isotopic inheritance, or if present expected/stated Luminescence OSL, IRSL • Quartz-derived ages have a higher rating than feldspar-derived ages • Single-grain or small aliquot • Homogenous sample; preferably aeolian or littoral, fluvial, glacifluvial sediments that are likely to have received sufficient exposure • Sample setting considered and accounted for; e.g.water-content history • Dose rate information and equivalent dose including errors described in source • Multiple and/or stratigraphically consistent ages All dating methods • Sample considered in situ, i.e. no post-depositional disturbance or reworking • Specified error margins • Precise ages: errors < 10% of age • Details of geological and stratigraphical setting given • Considered by original authors to be reliable have been entombed by (passive) ice (Mangerud and Svendsen, 2017 and others therein;Miller et al., 2017).We associate warm regional waters to thermophilous marine molluscs e.g.Arctica islandica, Mytilus edulis, Modiolus modiolus and Zirfaea crispata.These species of molluscs have been sampled around Svalbard and radiocarbon date to different periods in the Holocene.The occurrence of these shallow marine molluscs, which are "effectively" absent from Svalbard fjords during the 20th century, suggests warmer than present conditions and are indicators of the (marine) Holocene Thermal Maximum on Svalbard (Feyling-Hanssen, 1955;Salvigsen et al., 1992;Hjort et al., 1995;Salvigsen, 2002;Hansen et al., 2011;Farnsworth et al., 2017;Mangerud and Svendsen, 2017).
We classify radiocarbon ages of terrestrial plants that have been entombed by (passive) ice as indicators of ice-cover expansion (ICE; Miller et al., 2017).Contrary to ages associated to a "re-advance", these samples do not correspond to unique/specific glacier systems.Large populations of these ages sampled over a vertical range have been associated with regional snow-line lowering, however a sole age does not indicate a conclusive driving factor of ice-cover expansion (summer temperature, winter precipitation, wind direction or some combination).

The SVALHOLA database
The SVALHOLA database contains 1837 individual dates from over 722 discrete locations compiled from nearly 300 published sources (Fig. 4, Table S1-3).The spatial distribution is uneven and ages are predominantly from the western coast and fjords of Spitsbergen.Within the database, the lacustrine archive has the lowest density of ages (~10 %) followed by marine cores (~29 %).Samples dated from terrestrial archives make up the remaining population (~61 %).Over 90% of the dates in the SVALHOLA database are from radiocarbon analysis and roughly 50% of those samples are from marine organisms.In the database, there is a total of 104 ages from luminescence dating (~6 %) while there is a total of 46 ages (~3 %) from TCN dating.
Holocene radiocarbon ages from Svalbard are skewed toward the earlier half of the Holocene (Fig. 5).Roughly 60% of the SVALHOLA database date prior to 6.0 ka BP with nearly half of those ages falling between 11.5 -9.5 ka BP (Fig. 5).Roughly 20% of the radiocarbon ages within the SVALHOLA database date to within the last 2 ka BP.Although the Middle Holocene has comparatively fewer dates per 500year period, there is a relatively consistent distribution of ages, with a gentle decline 9 -2 ka BP (Fig. 5).There is a lower percentage of "likely reliable" (quality 1) radiocarbon ages versus less reliable ages in the Early Holocene.
There are a greater percentage of likely unreliable (quality 3) ages in both the TCN and luminescence populations compared to the radiocarbon ages.This is predominantly related to age precision and error ranges exceeding 10% of the age, especially for the younger half of the sample distribution.While the bulk of the luminescence ages fall within the Mid Holocene there is an inverse distribution of TCN ages, with the majority dating to the Late Holocene and the second largest population in the Early Holocene (Fig. 6).

Holocene Glacier Activity
Our results are compilations of records and ages that (in-)directly enhance our understanding of Holocene glaciers on Svalbard.Here we present a series of age-transects that reflect the Early Holocene deglaciation (rates and percentages) within Svalbard fjords.Additionally, we map the spatial and temporal distribution of glacier re-advances that punctuated this phase of deglaciation during the Early Holocene.By compiling glaciolacustrine and fjord records from Svalbard, we are able to reconstruct Mid Holocene glacier (in-)activity and identify where bodies of ice likely persisted.Furthermore, compilations of glacial lake and fjord records detail Late Holocene glacier expansion, complementing our mapped distribution of glacial landforms constrained to Late Holocene glacier fluctuations.

Retreat within Early Holocene fjord systems
By 12.0 ka BP the mouths of Svalbard fjords had already begun to deglaciate (Hormes et al., 2013;Hughes et al., 2016).Here we highlight six fjord system transects, marking the oldest Holocene ages (found in the SVALHOLA database), from fjord-mouth to fjord-head, to indirectly track the ice marginal retreat during the Early Holocene (Fig. 7).The oldest ages found at the heads of fjord systems range between 11.5 -9.1 ka BP and averages 10.5 ka BP for all fjord systems.Spatial comparisons of the ages suggest retreat rates within fjord systems ranged from roughly 10 km per 1 ka (St.Jonsfjorden) to as great as 12 km per 0.1 ka (Wijdefjorden).Normalized fjord lengths by percent suggest Svalbard fjords had become half-way deglaciated by 11.6 ± 0.2 ka BP and 75% ice-free by 11.3 ± 0.4 ka BP (Fig. 7).Generally, greater rates of Early Holocene retreat are found in longer fjord systems.

Early Holocene retreat punctuated by glacier re-advances
As Svalbard fjord systems became ice-free during the Early Holocene, tributary ice margins (from cirques, valleys and outlet glaciers) exhibited re-advances (Fig. 8).Reworked dateable material from the Early Holocene has been identified within and sampled from glacial deposits across Svalbard.Most often, this material is shell fragments resedimented in glacial (marine) diamictons suggesting glacier overriding and in some cases glaciotectonism (e.g.Salvigsen et al., 1990;Lønne, 2005;Farnsworth et al., 2018).The highest frequency of Early Holocene glacier re-advances (11 of 20 events) occurred between 11 -10 ka BP (Fig. 8).While re-advances associated with younger ages are located at the heads of fjord systems, older Early Holocene re-advances are identified near the outer extent and mouths of the fjords (Farnsworth et al., 2018;Larsen et al., 2018).The youngest Early Holocene glacier re-advances are found at the head of Wijdefjorden (9.1 ± 0.19 ka BP; Marks and Wysokinski, 1986) and Tempelfjorden (9.1 ± 0.14 ka BP; Forwick et al., 2010; Table S1).

Middle Holocene ice-free lake catchments and reduced glacier influence within fjord systems
Svalbard's glacial lacustrine sedimentary archives have been investigated in order to further reconstruct Holocene glacier activity (e.g., Bøyum & Kjensmo 1980;Svendsen and Mangerud, 1997;Røthe et al., 2015).Glacial lacustrine sedimentary archives provide a unique window into Svalbard's Middle Holocene glaciers during a time when glacier moraine records are absent.A compilation of glacial lake records extending through the Holocene suggests catchments received little or low minerogenic input which has been interpreted as reduced or ice-free catchments (Fig. 9).Lake records suggest lake catchments deglaciated between 10 -7 ka BP and remained ice-free through the Middle Holocene.As of yet, no (investigated) glacier lake record on Svalbard suggests active glaciers within the catchment through the Middle Holocene.
Glacial reconstructions from lake cores are only as robust as the sediment core chronologies.The distribution of ages within the SVAL-HOLA database exhibits Svalbard lake records are some of the highest concentrations of radiocarbon ages per point-location.However, the resolution of lake core chronology ranges widely as a result of core length and availability of datable material (Fig. 9).In lake sediment cores, a mixture of material has been used to constrain chronologies (Fig. 8).Early investigations using lakes to reconstruct palaeo-environmental conditions relied on radiocarbon dating bulk sediment (e.g.Hyvärinen, 1970), while modern studies target terrestrial plant macrofossils (de Wet et al., 2018) and occasional mollusc shells sampled just below the lacustrine-marine boundary in isolation basins (Svendsen and Mangerud, 1997;Schomacker et al., 2019).Recent studies have used crypto-tephra to constrain lake chronologies where terrestrial macrofossils are sparse (Fig. 8;D'Andrea et al., 2012;van der Bilt et al., 2017).Based on a collection of Svalbard's glacial lake records, composite lake cores average 2.4 m in length and are well dated, averaging 4.6 ages per meter (Fig. 9).
Marine sedimentary archives have also been investigated in Svalbard fjords to enhance Holocene glacier reconstructions (Fig. 10).While some marine records suggest glaciers retreated back to terrestrial margins during the Early Holocene and coasts remained ice-free through the Middle Holocene, other records suggest continuous (although reduced) input of outsized particles of ice rafted debris (IRD).Studies have suggested that location dependence can greatly influence whether IRD is glacier derived or originating from sea ice plucking beach sediments (Forwick et al., 2010;Joo et al., 2019).
Similar to lake cores, marine core chronology ranges widely as a result of core length and availability of datable material (Fig. 10).Holocene reconstructions from marine sediments often use mollusk shells or foraminifera for radiocarbon dating and constraining core  S1-S3 and at the end of this manuscript (full references are included in the supporting information; Data S1).
chronologies.Unlike lake records, crypto-tephra has not been utilized chronologically within Svalbard fjord records.Generally, Svalbard's Holocene marine records, exhibit lower chronological resolution than lake records.Composite Holocene fjord records average 4.9 m in length and average 1.8 ages per meter (Fig. 10).

Late Holocene glacier expansion
Starting c. 4.0 ka BP lake and fjord records across Svalbard begin to see increased sedimentation rates generally characterized by enhanced mineragenic input into lake basins and increasing IRD within fjord records (Figs. 9 & 10).Enhanced sediment transport from catchments to basins (between 4 -0.5 ka BP) is associated with widespread Late Holocene glacier expansion, referred to as the Neoglacial (Werner, 1993).Late Holocene glacier expansion is also reflected by glacial landforms associated with glacier growth and re-advance.Landforms and Fig. 5. Histogram of radiocarbon ages contained within the SVALHOLA database with dates presented with a 500-year bin size modified from 1000-year bin size presented by Hughes et al. (2016).Dates colored according to quality rating 1 (dark blue), 2 (blue) and 3 (light blue).The plot displays the same 1667 SVALHOLA dates with the centers for the age distributions stacked chronologically (black dots) and associated error bars (grey).Ages extend over 12.0 ka BP as the database includes all ages in which error margins fall within the Holocene (Table S1).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)sediment have been constrained in time where molluscs, whalebones or vegetation have been radiocarbon dated or TCN dating has been conducted on glacier-deposited boulders.Despite no geochronological constrains on glacial deposits dating to the Middle Holocene (8.2 -4.2 ka), there are abundant exposure ages as well as molluscs found within re-sedimented or overlain by glacial deposits of Late Holocene age (since 4.2 ka; Fig. 11).Roughly half of Late Holocene glacier re-advances are constrained by re-worked molluscs found within glacial deposits (Punning et al., 1976;Werner, 1993;Sharin et al., 2014;Farnsworth et al., 2017;Lovell et al., 2018).
Nearly half of the ages constraining Late Holocene glacier advances (in addition to the mollusc shells) are from over-ridden (in situ) vegetation (Baranowski and Karlén, 1976;Dzieriek et al., 1990;Furrer 1991;Humlum et al., 2005).Ages become more frequent in the last 2 millennia (Fig. 11).In addition to material that has been reworked or overridden by glaciers, plants that have been entombed in passive ice (cold-based, perennial snow patches and firn) have also been sampled and dated (Fig. 11).Over 40 samples of moss and vegetation preserved under cold-based ice have been collected from retreating modern ice margins in central Spitsbergen (Miller et al., 2017).While the oldest ages range back to nearly 4 ka BP, c. 80% of the samples are younger than 2.0 ka BP (Fig. 11).

Holocene sea level and environment
Not all ages within the SVALHOLA database directly reflect glacier activity.However, many of these ages have been used to enhance our understanding of Svalbard's Holocene glacier history by providing chronological envelopes for environmental conditions that relate to the behavior of the glacier systems.

Variability in Svalbard's glacioisostatic rebound
The dominant fingerprint of Holocene coastal geomorphology on Svalbard is flights of raised beaches and marine sediments.These raised beaches indicate previous shorelines that have subsequently been uplifted relative to local sea level as a result of glacioisostatic rebound (Forman et al., 1987;Bondevik et al., 1995).For the last 60 years, postglacial relative sea level curves have been constructed for regions all over Svalbard (Feyling-Hanssen, 1955;Birks, 1991, Blake, 1962;Forman et al., 2004;Schomacker et al., 2019;Fig. 12).The variability in uplift exhibited in Svalbard's relative sea level curves is a result of glacial history and behavior.Svalbard's relative sea level history is a product of ice-cover thickness, duration, timing of deglaciation and in some cases, subsequent ice expansion (Ingólfsson and Landvik, 2013;Fjeldskaar et al., 2018).Elevations of postglacial marine limits (where curves have been developed) range from slightly under 20 m a.s.l. in Southern Spitsbergen to greater than 100 m a.s.l. on Kong Karls Land in Eastern Svalbard (Fig. 12).
Holocene uplift rates have been derived from radiocarbon dated whalebones, driftwood and shells found on or in raised beaches (Salvigsen, 1981;Häggblom, 1982;Forman et al., 2004).While there is large amount of variability, the Early Holocene exhibits the greatest rates of isostatic uplift recorded in the last 11.7 ka BP.For some locations, minimum uplift rates derived from raised marine shorelines suggest 10 -25 meters of uplift per 1.0 ka during the Early Holocene on Svalbard (Salvigsen, 1981;Salvigsen and Österholm, 1982;Forman et al., 2004).The ages of raised marine shorelines suggest glacioisostatic uplift rates decline throughout the Middle Holocene and equate to c. 5 m per 1.0 ka (Bondevik et al., 1995;Forman et al., 2004).In several locations on the northern and western coasts of Svalbard, regression ceased and Middle Holocene transgressions occurred, as interpreted through a combination of shoreline morphology and geochronology (Forman, 1990;Forman and Ingólfsson, 2000).Through the Late Holocene, dated raised marine shorelines indicate a further decrease in relative uplift rates (Forman et al., 2004).Generally, it is believed Late Holocene relative sea level was regressive around Svalbard (Bondevik et al., 1995;Forman et al., 2004;Sessford et al., 2015).It is unknown to what extent Neoglacial glacier expansion influenced relative sea level in Svalbard during the Late Holocene as eustatic sea level is suggested to have out-paced relative land uplift along Spitsbergen's western coast (Forman et al., 2004;Fjeldskaar et al., 2018).

Holocene driftwood
Radiocarbon ages of driftwood make up roughly 10 % of the SVA-LHOLA database and 35 % of the terrestrial ages.The occurrence of driftwood found on Arctic shorelines has been used as a proxy for semi-Fig.6. Histogram of cosmogenic nuclide (purple) and luminescence (orange) ages contained within the SVALHOLA database with dates presented with a 500-year bin size modified from 1000-year bin size presented by Hughes et al. (2016).Dates colored according to quality rating 1 (dark purple/orange), 2 (purple/orange) and 3 (light purple/orange).The plot displays the same ages from the SVALHOLA database with the centers for the age distributions stacked chronologically (colored circles) and associated error bars (grey) excluding the likely unreliable quality 3 ages.Ages extend over 12.0 ka BP as the database includes all ages in which error margins fall within the Holocene (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)permanent sea-ice extent (Häggblom, 1982;Funder et al., 2011;Nixon et al., 2016;Hole and Macias-Fauria, 2017).Where sea-ice cover is too low no driftwood arrives, while with multiyear sea ice, driftwood is shielded from the shorelines (Funder et al., 2011).The frequency of radiocarbon dated driftwood from Svalbard's raised marine shorelines exhibits an increasing rate of arrival between 12.0 -10.5 ka BP which is followed by a stepped and variable decline in occurrence to 9.0 ka (Fig. 13).Peak driftwood arrival occurs between 11.0 and 10.5 ka BP (Fig. 13).The presence of driftwood on raised Middle Holocene shorelines remains lower than the peak occurrence from the Early Holocene, yet suggests consistent arrival of material from roughly 9.0 to 6.5 ka BP.This interval is followed by a steady but variable arrival of driftwood from 6.5 to 4.2 ka BP (Fig. 13).During the first half of the Late Holocene driftwood occurrence remains low (Fig. 13).The period between 2.5 -2.0 ka BP marks the Holocene minimum in driftwood arrival and matches the occurrence rate seen during the end of the Younger Dryas Plots present retreat length (km) and percentage retreat (%) over time separately for each fjord system and comparatively for all fjord systems.Fjord length calculations are based on the distance between database ages rather than modern fjord dimensions.Thus, fjord lengths are in some cases underrepresented as ages do not reach modern fjord heads while other fjord lengths are overrepresented as ages are found in fjord valleys, presently located kilometers inland.period (12.0 -11.5 ka BP; Fig. 13).A sharp increase in the number of dated driftwood is exhibited during the last two millennia.The driftwood count from the most recent 500 years (bin 0.5 ka BP to modern) has been excluded given the general lack of studies targeting radiocarbon dated driftwood accumulated at or slightly above the modern shorelines.Furthermore, the occurrence of Holocene driftwood on raised marine shorelines is influenced by preservation, and therefore may as well be influenced by Middle and Late Holocene transgressions.

Thermophilous marine species
Several species of molluscs are suggested to indicate "warm regional waters" around Svalbard at different time intervals in the Holocene based on their modern habitats (Feyling-Hanssen, 1955;Salvigsen et al., 1992).These thermophilous molluscs that once inhabited Svalbard can be found preserved in raised marine sediments of Early-Middle Holocene age (Fig. 13; Feyling-Hanssen, 1955;Salvigsen et al., 1992;Hjort et al., 1995;Salvigsen, 2002;Blake, 2006;Hansen et al., 2011;Farnsworth et al., 2017;Mangerud and Svendsen, 2017).The earliest radiocarbon ages of the warm water species Mytilus edulis date slightly before 11.0 ka BP and suggest the sea temperatures around Svalbard were roughly 2 °C warmer than present (Fig. 13; Mangerud and Svendsen, 2017).Additionally, the occurrence of Zirfaea crispata and Arctica islandica around Svalbard has been used to suggest shallow ocean temperatures peak between 10.0 -9.2 ka BP and were at least 6°C warmer than present, framing the marine Holocene thermal maximum (HTM; Mangerud and Svendsen, 2017 and references therein).Thermophilous molluscs persist in Svalbard waters throughout the Middle Holocene.A slight decline in the occurrence of thermophiles suggests a short-lived cooling following the marine HTM between 9 and 8.2 ka BP.However, ocean temperatures still remained roughly 2°C warmer than present based on the persistence of Mytilus edulis (Mangerud and Svendsen, 2017).Both Mytilus edulis and Modiolus modiolus remained in Svalbard waters into the Middle Holocene suggesting ocean temperatures were c. 4 °C warmer than present 8.0 -6.5 ka BP (Salvigsen et al., 1992;Salvigsen, 2002;Blake et al. 2006;Mangerud and Svendsen, 2017).The occurrence of thermophilous molluscs decreases through the Middle Holocene and tapers off entirely at the start of the Late Holocene, when reconstructed ocean temperatures reach those comparable to modern (Fig. 13; Mangerud and Svendsen, 2017).In addition to the two Mytilus edulis dating c. 3.7 ka BP, it appears the species may have returned to inner Isfjorden based on a single young sample dating 0.9 ka BP (Fig. 13; Samtleben, 1985;Mangerud and Svendsen, 2017).
Warm waters are not just reconstructed from the shallow inner fjords-systems of Svalbard.Marine micropaleontology records from outer fjords and regional waters, also suggest warm conditions prior to the onset and throughout the Early Holocene (Hald et al., 2004;Ślubowska-Woldengen et al., 2007;Skirbekk et al., 2010;Rasmussen et al., 2012;Łącka et al., 2015;Allaart et al., 2020).Warm Atlantic water is associated with the dominance of N. labradorica and relative decline in N. pachyderma (sinistral) found within the Early Holocene fjord systems (Rasmussen et al., 2012).

Discussion
We begin by refining our understanding of ice cover during the transitions from Late Pleistocene to Early Holocene in comparison to previous reconstructions (Mangerud and Landvik, 2007;Hormes et al., 2013;Hughes et al., 2016;Hogan et al., 2017).We then discuss evidence of Holocene glacier activity from the marine, terrestrial and lacustrine archives for each stage: Early (11.7 -8.2 ka BP), Middle (8.2 -4.2 ka BP) and Late (4.2 ka BP to present) Holocene.Following each Holocene stage, we discuss reconstructed environmental conditions which likely influenced glacier behavior.We complete our discussion by considering the key drivers of Holocene glaciers on Svalbard and how ice margins behaved during cooling events and warming events.Despite certain micro-climatic variability, we choose not to subgroup data geographically, but generally discuss ages for the entire Svalbard region.In some cases, we refer to more detailed locations e.g. the mouth of a fjord, relative to the tributary or head of the fjord system.(Røthe et al., 2015), Kb = Kløverbladvatnet (Schomacker et al., 2019) and L = Linnévatnet (Svendsen and Mangerud, 1997;Svendsen and Mangerud, 1997;Snyder et al., 2000).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Bartels et al., 2018), Wdf = Woodfjorden (Bartels et al., 2017), Kf = Kongsfjorden (Skirbekk et al., 2010), Bf = Billefjorden (Baeten et al., 2010), Tf = Tempelfjorden (Forwick et al., 2010), VMf = Van Mijenfjorden (Hald et al., 2004), VKf = Van Kuelenfjorden (Kempf et al., 2013), Sf = Storfjorden (Nielsen and Rasmussen, 2018).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Late Pleistocene -Holocene transition
We recognize the importance of constraining the extent of glacier cover during the transition from Late Pleistocene to the Holocene.While timeslices give an approximate distribution of glacier cover over a region during a window in time, they often over simplify critical information (e.g., ice thickness and dynamics) and extrapolate between sparse data.Our dataset generally indicates the lack of ice (e.g., deglaciation or ice free conditions), rather than the actual distribution (e.g., ice marginal positions).However, where ample geochronological data is tied with morphology or stratigraphy it is possible to develop a map representing glacier cover, if spatial scale allows.We feel any attempt to map glacier distribution at regional scale for Svalbard, limited by constraining data (geochronology and morphology) is likely an overrepresentation of limited knowledge.We choose to limit our timeslices reconstructions to only target the maximum and minimum range of ice cover during the Holocene on Svalbard (Fig. 14A).
Maximum ice cover since 12.0 ka BP was likely around 12.0 ka BP, as residual SBSIS still covered the majority of Svalbard (Fig. 14).The exact ice extent is not well understood given the limited number of ice marginal positions and sparse data throughout the region.Additionally, it is unknown how tightly the deglaciation ages from marine sediment cores and raised marine sediments constrain the actual deglaciation as these ages are minimum constraining values and could potentially reflect conditions centuries after the actual deglaciation (Hald et al., 2004;Larsen et al., 2018).Reconstructions of ice margins during the transition from Late Glacial to Early Holocene are challenged by potential deglaciation ice-shelves that may have limited the formation of ice-marginal landforms (Furze et al., 2018;Farnsworth, 2018).Based on ages found in the SVALHOLA database we suggest the most-credible ice cover timeslice from DATED-1 likely over-estimates glacier cover on Svalbard at the transition from Late Pleistocene to Early Holocene (Fig. 14A; Hughes et al., 2016).
The same is true for comparisons to the regional reviews by Hormes et al. ( 2013) and Hogan et al. (2017).While there is minimal evidence of ice loss between 12.0 -11.7 ka BP (Fig. 14A), by 11.0 ka BP it seems the bulk of fjords and land submerged under postglacial relative sea level are likely ice free (Fig. 14B).Fig. 11.Distribution of Late Holocene glacier re-advances associated with glacial landforms or deposits (marked by colored squares).Late Holocene snow-linelowering (marked by colored circles) indicates where dead vegetation identified at passive ice margins has been radiocarbon dated.Constraining ages occur from 4 -0 ka BP with nearly 50% of Late Holocene glacier re-advances dating to between 1 -0.5 ka BP.Dates were compiled from references presented in Tables S1-S3 and at the end of this manuscript (full references are included in the supporting information; Data S1).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)4.2.Early Holocene (11.7 -8.2 ka BP)

Early Holocene glaciers
Early Holocene glaciers on Svalbard undergo extensive ice-mass-loss (Fig. 7 & 14).Conceptually, deglaciation continued in a time-transgressive manner, characterized by initial retreat through the fjords and subsequently in the fjord-valleys and terrestrial realms (Ingólfsson and Landvik, 2013;Gilbert et al., 2018).Deglaciation is observed across marine, terrestrial and lacustrine records (Fig. 9,10,12 & 14).Fjords, lake basins and shorelines become ice-free allowing for the deposition of datable material (Fig. 13).Marine sediment cores from fjords in western and northern Svalbard exhibit high, yet diminishing sedimentation rates with persistent outsized particles interpreted as IRD from retreating marine ice margins (Hald et al., 2004;Forwick and Vorren, 2009;Flink et al., 2017;Bartels et al., 2017Bartels et al., , 2018;;Allaart et al., 2020).Similarly, lacustrine records from Svalbard often exhibit a high minerogenic fraction with sedimentation rates that taper off through the Early Holocene (Snyder et al., 1994;Svendsen and Mangerud, 1997;Røthe et al., 2015;van der Bilt et al., 2016;de Wet et al., 2018;Røthe et al., 2018;Schomacker et al., 2019).Records have been interpreted to indicate most marine-terminating glaciers retreated onto land, while cirque glaciers greatly diminished in size and in some cases disappeared completely (Svendsen and Mangerud, 1997;Snyder et al., 2000;Forwick and Vorren, 2009;de Wet et al., 2018).On northeastern Nordaustlandet, shells sampled from within a thrusted debris band located 6 km inside the modern ice margin of a major outlet glacier date to 10.3 ± 0.49 ka BP and suggest that the fjord system had retreated to at least this position early in the Holocene (Blake, 1989).Thus, by 10 ka BP most tidewater glacier margins are located well inside of their late Holocene extents.
However, glacier activity during the Early Holocene is complex and does not solely indicate consistent ice-marginal retreat.Despite the progressive deglaciation, evidence from glaciers of varying size found across Svalbard suggests asynchronous ice margin re-advance(s) (Fig. 8 & 13; Salvigsen et al., 1990;Mangerud et al., 1992;Ronnert and Landvik, 1993;Brückner et al. 2002;Eitel et al., 2002;Lønne, 2005;Forwick et al., 2010;Farnsworth et al., 2017Farnsworth et al., , 2018;;Larsen et al., 2018).Although the exact magnitudes of the re-advances are unknown, ice margin extent is often several kilometers distal to Late Holocene glacier maxima (Lønne, 2005;Farnsworth et al., 2018).Generally, the timing of the re-advances seems to follow the time transgressive deglaciation (Landvik et al., 2014;Farnsworth et al., 2018).The earliest re-advances have been identified near the mouths of the fjords while younger Early Holocene re-advances are found in the inner tributaries and heads of fjord systems (Larsen et al., 2018).The ultimate glacier re-advance (c.9.1 ± 0.19 ka BP) identified in the Early Holocene redeposited shells up to 180 m a.s.l. in inner Wijdefjorden (c.6.5 km inside and c. 100 m above of the present glacier margin; Marks & Wysokinski 1986;Klysz et al., 1988; Table S1).Although the extent of the re-advance interpreted from these deposits is unknown, the data suggests not only had the palaeo-outlet glacier undergone significant Early Holocene retreat, it re-advanced to a marginal position unmatched at any point in the Late Holocene.There are still numerous pre-Middle Holocene moraines that have been identified in marine and terrestrial environments, that remain poorly dated (Salvigsen and Österholm, 1982;Forwick et al., 2010;Henriksen et al., 2014;Røthe et al., 2015;Flink et al., 2017;Farnsworth et al., 2018;Flink and Noormets, 2018;Dowdeswell et al., 2020).

Early Holocene environment
Environmental reconstructions from marine archives suggest early and marked warm regional conditions driven by the incursion of Atlantic waters around Svalbard prior to 11 ka BP (Hald et al., 2004;Forwick and Vorren, 2009;Rasmussen et al., 2012;Mangerud and Svendsen, 2017).Evidence from marine microfossils suggests warm regional conditions prior to the onset and throughout the Early Holocene (Hald et al., 2004(Hald et al., , 2007;;Ślubowska-Woldengen et al., 2007;Skirbekk et al., 2010;Łącka et al., 2015).Reconstructions of sea-ice extent based on the sea-ice biomarker IP 25 (Müller and Stein, 2014;Bartels et al., 2018;Allaart et al., 2020) and IRD flux (Hald et al., 2004;Forwick and Vorren, 2009), suggest a decline in sea-ice extent around the onset of the Early Holocene c. 12-11 ka BP.Additionally, there is a high and variable accumulation of driftwood during the Early Holocene (Fig. 13).The initial peak in Early Holocene driftwood could be related to sea-ice cover transitioning from permanent to semi-permanent at the onset of the Holocene as well as the increased availability of deglaciated shorelines as catchment for driftwood (Fig. 13 & 14;Häggblom, 1982).
The records of raised marine mollusc shells identified in Early Holocene sediments across the coastal regions of Svalbard corroborates with the environmental reconstructions from fjord records.Reviews of thermophilous marine molluscs found around Svalbard also indicate early and exceptionally warm conditions within fjords (Salvigsen et al., 1992;Salvigsen, 2002;Blake, 2006;Mangerud and Svendsen, 2017).The occurrences of Zirfaea crispata and Arctica islandica in the Svalbard raised marine record only between 10.0 -9.2 ka BP suggests that the maximum ocean temperatures occurred soon after the onset of the Holocene and prior to peak summer insolation (Mangerud and Svendsen, 2017).
Recent Early Holocene studies targeting ice-free terrestrial and lacustrine archives suggest the terrestrial environment is in phase with marine temperature reconstructions.Alkenone and hydrogen isotope records from low-altitude coastal lakes suggest early warm and moist conditions with peak Holocene temperatures reaching 7 °C warmer than today, c. 10 ka BP (Balascio et al., 2018;Gjerde et al., 2017;van der Bilt et al., 2018van der Bilt et al., , 2019)).Warm conditions have also been documented from sedaDNA targeting vascular plants within lake sediment records (Alsos et al., 2015;Voldstad et al., 2020).Occurrence and diversity of thermophilous plant species like Empetrum nigrum, Arnica angustifolia and Arabis alpine indicate that during the Early Holocene, the ice-free terrestrial realm on Svalbard was warmer than present.Although the majority of the terrestrial landscape is presumed to still be evacuating Fig. 12. Plot exhibits variability in postglacial relative sea level from around Svalbard.Map indicates the distribution of sea level curves according to (Forman et al., 2004 and references therein;Sessford et al., 2015;Schomacker et al., 2019).Five relative sea level curves have been highlighted in bold, selected from different key sites around Svalbard, exhibiting unique variability in postglacial sea level; P = Prins Karls Forland, H = Hornsund, B = Bangenhuk, E = Edgeøya (Humla) and K = Kong Karls Land.Curves constructed from data present in Table S1 with full references located in Data S1).SBSIS-ice in the Early Holocene, records of terrestrial plant diversity and lake temperatures suggest that ice-free terrain also exhibited early and exceptional warmth.

Middle Holocene glaciers
We have no exact spatial constraint on any Svalbard ice margins during the Middle Holocene.It is assumed that if glaciers survived the Fig. 13.Plot of the 930 radiocarbon ages (excluding lake and marine records) from the SVALHOLA database with the centers for the age distributions stacked chronologically (black dots) and associated error bars (grey) modified from Fig. 5.The Younger Dryas period is marked with a blue column while the Early and Late Holocene is divided by a grey column delimiting the Middle Holocene (Cohen et al., 2018).Specific radiocarbon ages from the database have been highlighted based on palaeoglaciological or environmental associations.Dateable material reworked in glacial sediments (blue); thermophilous marine molluscs (red) and ice-entombed moss (green) are represented by enlarged colored circles with error whiskers.Histogram plot at base indicates the number (total n=170) of radiocarbon dated driftwood samples from Svalbard within a 500-year bin size (modified from Dyke et al., 1997).The youngest bin (modern -0.5 ka BP) has been excluded due to lack of studies focusing on driftwood from modern shorelines.Data available in Tables S1-S3.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Middle Holocene, they were more reduced than any Late Holocene glacier extent on Svalbard (Mangerud and Landvik, 2007;Baeten et al., 2010).To our knowledge, there is no geomorphological evidence (e.g., end moraines) identifying any glacier margins, stand-stills or re-advances during this period (Fig. 13).At the mouth of Hornsund, terrestrial vegetation thrusted up into tidewater glacier ice suggests the marine terminus was terrestrial grounded prior to 8.0 ka BP (Oerlemans et al., 2011).
At present, there are over 850 km of marine-terminating glacier margins in Svalbard (Blaszczyk et al., 2009).However, it has been speculated if tidewater glaciers survived the Middle Holocene.There are two marine stratigraphical studies that have suggested Middle Holocene glacier activity at the mouth of Isfjorden and a bay on western Nordaustlandet based on fluctuating rates of minerogenic sedimentation (Forwick and Vorren, 2007;Kubishcta et al., 2011).However, it remains inconclusive if fluctuations in marine sedimentation from these sites relates to actual glacier re-advances or rather enhanced runoff from proximal glacier systems, submarine mass failures, or enhanced sea ice rafting.Most marine sediment cores reveal a decline in sedimentation rates, and IRD dissipates in many of Svalbard's fjord systems (Forwick and Vorren, 2009;Skirbekk et al., 2010;Kempf et al., 2013;Nielsen and Rasmussen, 2018).However, ice-rafted debris has been observed through the entire Holocene in sediment cores from several Svalbard fjords, including Van Mijenfjorden, Billefjorden, Tempelfjorden and Wahlenbergfjorden (Hald et al., 2004;Forwick andVorren, 2009, 2010;Baeten et al., 2010;Flink et al., 2017;Bartels et al., 2018).Ice-rafted debris is a strong indicator of marine terminating glaciers within a fjord but has also been associated with sea ice, serving both as a raft for beach sediments as well as a (semi-) permanent restraint limiting ice rafting by icebergs (Forwick and Vorren, 2009;Joo et al., 2019).Assuming that Holocene IRD within Svalbard fjords relate to tidewater glacier systems, the IRD records indicate that Svalbard glaciers not only survived the entire Holocene interglacial, but several Svalbard glaciers remained at their fjord-heads through the Holocene.Using IRD as a proxy for tidewater glaciers maybe less accurate within some shallow, inner Svalbard fjords (e.g., Dicksonfjorden) where sea ice appears to play a larger role in the rafting of outsized clasts to the marine archive (Joo et al., 2019).
It remains difficult to determine the Holocene minimum ice extent, and we can only speculate when glaciers on Svalbard reached a Holocene minimum (Fig. 15).While some lake sediment studies targeting cirque glaciers suggest that catchments were ice-free through the Middle Holocene, lake studies have yet to target larger lake basins corresponding to large glacier systems.Presumably, the Holocene glacial minimum occurred in the Middle Holocene, sometime between 8.0 -6.0 ka BP.The glacial minimum certainly occurred after the marine thermal maximum (10.0 -9.2 ka BP), potentially lagging the peak in Fig. 14.IBCAO map of Svalbard with polygons reflecting ice cover during the transition from Late Glacial into the Early Holocene based on constraining deglacial ages from the SVALHOLA database.A) displays regional ice cover at around 12.0 ka BP and at 11.7 ka BP compared to the most-credible DATED-1 12.0 ka BP timeslice (Jakobsson et al., 2012;Hughes et al., 2016).B) indicates reconstructed ice marginal retreat between 11.7 and 11.0 ka BP.Northernmost ice cover polygon over the Seven Islands reflects reconstructed ice cover that existed until the Early Holocene (but lacks from other reconstructions due to resolution).Timeslice footprints represent the approximate extent of ice-cover during a period in time, yet take no account for glacier behavior or ice thickness (Hormes et al., 2013).
ocean temperatures and summer insolation on a millennial time scale.
dominant or the only source of IRD during the Early and Middle Holocene (Fig. 15).Based on the size of the Reindalen and Kjellströmdalen valleys, they seem to have held larger glacier systems in their catchments and may have more efficiently preserved their residual SBSIS ice.

Middle Holocene environment
Data from the marine archive suggests a general decline in ocean temperatures during the Middle Holocene.Cooling of regional waters during this period is indicated by marine microfossils (Hald et al., 2004(Hald et al., , 2007;;Ślubowska-Woldengen et al., 2007;Skirbekk et al., 2010;Łącka et al., 2015).Reconstructions of sea-ice extent derived from IP 25 biomarker records spanning the Middle Holocene suggest a progressive increase in sea-ice cover starting around 8.0 ka BP (Müller and Stein, 2014).
Raised Middle Holocene marine sediments and the persistence of Mytilus edulis and Modiolus modiolus therein suggest Atlantic water continued to reach Svalbard, and ocean temperatures were c. 4 °C warmer than present between 8.0 -6.5 ka BP (Salvigsen et al., 1992;Salvigsen, 2002;Blake, 2006;Mangerud and Svendsen, 2017).This period of stable warm ocean conditions is followed by a gradual decrease in temperatures until the end of the Middle Holocene, when sea surface temperature reached those comparable to modern (Hald et al., 2004;Mangerud and Svendsen, 2017).
Evidence from terrestrial and lacustrine archives suggests the Middle Holocene environment on Svalbard was warmer than present, yet conditions had cooled since the peak in temperatures reconstructed from the Early Holocene.The Middle Holocene is characterized by a gradual, step-wise cooling, with dry conditions and summer lake temperatures 2 -4°C warmer than present (Bakke et al., 2018;Røthe et al., 2018).Vascular plant reconstructions from sedaDNA within lake sediment records present similar conditions, exhibiting cooling during the later Middle Holocene (Alsos et al., 2015;Voldstad et al., 2020).

Late Holocene glaciers
The standard perspective that the (Late) Holocene glacial maximum marked the culmination of the LIA around the turn of the 20th Century AD (Svendsen and Mangerud, 1997;Snyder et al., 2000) is increasingly being challenged.Svalbard glaciers re-advanced throughout the Late Holocene, during the Neoglacial (Furrer et al., 1991;Werner, 1993;Reusche et al., 2014;Sharin et al., 2014;Røthe et al., 2015;van der Bilt et al., 2016;Philipps et al., 2017) and during the LIA (Baranowski and Karlén, 1976;Dzieriek et al., 1990;Svendsen and Mangerud, 1997;Snyder et al., 2000;Humlum et al., 2005;de Wet et al., 2018).Where Neoglacial ice margins have been identified, they are often on the order of tens of meters outboard LIA margins (Werner, 1993;Sharin et al., 2014).Observations of glaciers during the early 1900s describe Svalbard glaciers proximal to their Late Holocene maximum extents (Isachsen, 1915).However, this does not suggest anything about the timing of glacier advance, but rather that the glaciers had started to retreat.
The majority of Holocene glacier re-advances are constrained by radiocarbon dates from material re-deposited in or overlain by glacial deposits during the Late Holocene (Fig. 11 & 16).Most Late Holocene glacier re-advances (dated by overridden vegetation or mollusc shells) date to the early (between 1.0 -0.5 ka BP), not the late LIA (Fig. 11).Dates constrain the maximum age of glacier re-advance during the Neoglacial or early LIA when glaciers grew to their Late Holocene maximum extent (Baranowski and Karlén, 1976;Dzieriek et al., 1990;Furrer et al., 1991;Humlum et al., 2005).Several additional studies indicate Late Holocene glacier maxima occurred even earlier in the Neoglacial (Werner, 1993;Sharin et al., 2014;Lovell et al., 2018).
Despite criticism of the method (e.g.Osborn et al., 2015), lichenometry studies suggest glaciers across Svalbard have experienced several phases of glacier advance, followed by moraine stabilization during the last 2.0 ka BP (Fig. 11;Werner, 1993).Cosmogenic exposure dating of moraine ridges has been conducted on several Late Holocene glacier forelands as a means of testing whether the LIA was the most extensive glacial event of the Late Holocene (Fig. 11; Reusche et al., 2014;Philipps et al., 2017).The TCN studies have effectively supported earlier works, corroborating there have been numerous phases of glacial re-advance during the Late Holocene and conclude that the LIA is not the largest glacial event in most locations.
Further evidence of episodic glacier expansion during the Late Holocene is derived from the technique of radiocarbon dating ice-entombed plants over a range of elevations to reconstruct snow-linelowering throughout a region (Miller et al., 2017).Results from a study presenting over 40 radiocarbon ages from central Spitsbergen suggest that there were at least four phases of widespread ice cover expansion resulting in a general snow line lowering between 2.0 ka BP and 0.5 ka BP prior to the LIA-ice expansion (Miller et al., 2017).They further demonstrate that ice expansion occurred as early as 4.0 ka BP (Fig. 11; Miller et al., 2017).Findings align with results from active ice margins however, this work has closely relied on the assumption that Arctic glaciers are solely controlled by summer temperature, overlooking the potential role of precipitation, shifts in storm-track or wind (Humlum et al., 2005;Wickström et al., 2020).
The record of ice entombed plants places ice-core stratigraphy studies in perspective (Isaksson et al., 2005;Divine et al., 2011;Miller al., 2017).Previous ice core sample locations have targeted sites with continuous, undisturbed snow accumulation, which have extended back approximately 1 ka BP (Isaksson et al., 2003(Isaksson et al., , 2005;;Grinsted et al., 2009;Divine et al., 2011).However, ice-entombed vegetation ages suggest glacier cover nearly as old as the Middle Holocene, albeit lacking record continuity.Fragmented, albeit old records of precipitation may exist in Svalbard's cryostratigraphy, but have yet to be targeted or identified.Given the potentials of (crypto-)tephra records on Svalbard, older ice core chronologies may be possible despite lacking annual layer counts from the modern surface (Isaksson et al., 2005;Kekonen et al., 2005;Wastegård and Davies, 2009).Furthermore, the identification, mapping and investigation of old Holocene ice may provide information about the glacial minimum, shifts in precipitation patterns as well as processes related to glacier expansion.
Late Holocene glacier expansion is also interpreted from fjord and lacustrine sedimentary archives.Sedimentation rates in Svalbard fjords increase through the Late Holocene (Fig. 10; Hald et al., 2004;Forwick and Vorren, 2009;Baeten et al., 2010;Kempf et al., 2013;Łącka et al., 2015;Streuff et al., 2017;Flink et al., 2017;Bartels et al., 2018), interpreted to mean increased glacial activity in their catchments.Also, Late Holocene proglacial lake records generally suggest increasing glacial influence, characterized by rising rates of sedimentation and transitions from organic-rich to minerogenic strata (Fig. 9 &16).Some lake sediment records suggest episodic increases in glacial activity within catchments starting as early as c. 4 ka BP (Hajeren;van der Bilt et al., 2016) and prior to 3.0 ka BP (Kløsa and Linnévatnet; Røthe et al., 2015;Svendsen and Mangerud, 1997).In Gjøavatnet (1 m a.s.l.), located in northwestern Svalbard, the transition from organic-rich to glacial sedimentation does not occur until c. 1 ka BP.All glacial lake records show enhanced glacial influence during the last millennium (Fig. 11).A regional glacier equilibrium line altitude (ELA) has been reconstructed from the Kløsa lake sediment record in northwestern Svalbard suggesting variable but declining ELA over the last 4 ka BP (Røthe et al., 2015).
Several glacier re-advances constrained to the Neoglacial -LIA have been characterized as surges based on size, landsystem, extent of glacial deformation and preservation of landforms (related to rapid ice advances) corresponding to associated ice-margins (Plassen et al., 2004;Ottesen et al., 2008;Kristensen et al., 2009;Kempf et al., 2013;Farnsworth et al., 2016Farnsworth et al., , 2017;;Flink et al., 2017;Lovell and Boston, 2017;Lyså et al., 2018;Lovell et al., 2018;Aradóttir et al., 2019).While the most extensive Late Holocene glacial deposits have been associated with surge-type behavior during, or at, the culmination of the LIA (Schomacker and Kjaer, 2008;Kristensen et al., 2009;Kempf et al., 2013;Flink et al., 2015;Lyså et al., 2018) an increasing number of studies have identified both complete and fragmented moraine ridges outboard of the LIA maxima (Werner, 1993;Sletten et al., 2001;Reusche et al., 2014;Sharin et al., 2014;Philipps et al., 2017;Larsen et al., 2018).(Reusche et al., 2014;Philipps et al., 2017).Shaded bars at top represent a schematic view of Late Holocene glacial activity reconstructed from lake sediments of four proglacial lakes from western Spitsbergen (Gjøavatnet, Hajeren, Kløsa and Linnévatnet).Blue boxes represent periods of glacial activity (dark blue suggest enhanced glacial activity), red boxes suggest no glacier was present in the catchment or glacial activity was greatly reduced; gray boxes indicate periods of reduced organic matter accumulation (Gjøavatnet; modified from de Wet et al., 2018).Light blue boxes indicate periods of glacier growth interrupted by moraine stabilization (white boxes) reconstructed with lichenometry (Werner, 1993).Data available in Tables S1 & S2.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Late Holocene environment
The combination of decreasing summer insolation (Berger and Loutre, 1991;Laskar et al., 2004) and the progressive cooling of regional waters draws Svalbard into the Neoglacial period (Fig 13; (Skirbekk et al., 2010;Rasmussen et al., 2014;Łącka et al., 2015;Svendsen and Mangerud, 1997).The Neoglacial is characterized by gentle decline in temperatures initiated before the onset of the Late Holocene, c. 5.0 ka BP, superimposed by a series of episodic (multidecadial to multicentenial) steps in deteriorating climate throughout the North Atlantic (Bradley and Bakke, 2019).
The period 2.5 -2.0 ka BP marks the Holocene minimum in driftwood arrival and matches the occurrence rate seen during the end of the Pleistocene (Fig. 7).It is unclear how well driftwood arrival through the Late Holocene reflects sea-ice cover around Svalbard (Fig. 7).However, it is possible the Holocene minimum in driftwood arrival relates to persistent or semi-permanent land-fast ice, minimizing the transport of and catchment for driftwood accumulation (Funder et al., 2011).The increase in driftwood arrival exhibited through the last 2 ka of the Holocene record resembles the increase seen in the first half of the Early Holocene (Fig. 7).IP 25 derived sea-ice reconstructions suggest variability in sea-ice cover during the Late Holocene associated with sporadic warm sea surface temperatures (Müller et al., 2012;Sarnthein et al., 2003;Jernas et al., 2013;Pawłowska et al., 2020).Brine formation is believed to reflect sea-ice cover above basins and a Late Holocene record from Storfjorden suggests episodic periods of intense brine production separated by periods of reduced brine formation (Rasmussen and Thomsen, 2014).
The re-occurrence of Mytilus edulis at 0.9 ka BP (in Dicksonfjorden) coincides with the "Medieval Warm Period" and may suggest warm Atlantic waters were more widespread around Svalbard (Fig. 16;Lamb, 1965;Mann et al., 2009).The mapped distribution of these warm water molluscs around Svalbard from Mangerud and Svendsen (2017) suggests the species thrive in the inner shallow branches of fjord systems with nominal glacial influence (e.g., Dicksonfjorden Woodfjorden and outer Billefjorden; Salvigsen et al., 1992;Salvigsen, 2002;Blake, 2006;Mangerud and Svendsen, 2017).Thermophilous molluscs seem to generally reflect surface water conditions around Svalbard, however, may not necessarily represent the large-scale oceanographic condition that existed on the continental shelf or towards the east of Spitsbergen.
Lacustrine records from non-glacial lakes (Hakluytvatnet, Skardstjørna and Jodavatnet) suggest decreasing productivity, declining plant species diversity as well as increased Late Holocene minerogenic input due to increased runoff (Alsos et al., 2015;Gjerde et al., 2017;Voldstad et al., 2020).Additionally, a 1.8 ka alkenone based (summer) temperature record from Kongressvatnet, located in central western Spitsbergen, suggests the LIA was relatively mild in the region, implying precipitation may-be the key driver behind glacier expansion (D'Andrea et al., 2012).While it is important to highlight our limited knowledge of past precipitation, there is ample evidence suggesting that air and ocean temperatures were relatively cool and favored glaciers during the late Neoglacial and LIA (Divine et al., 2011;Bartels et al., 2017;Røthe et al., 2018;van der Bilt et al., 2018;Balascio et al., 2018;Luoto et al., 2018).The Kongressvatnet record however, does highlight that the LIA was not the main event of the Late Holocene rather a late phase of a generally cool Neoglacial period.
As previously noted, sea-ice conditions during the Neoglacial and LIA are characterized by increasing but variable sea-ice extent around Svalbard (Fig. 12; Müller et al., 2012;Jernas et al., 2013;Müller andStein, 2014Bartels et al., 2017;Allaart et al., 2020).Although increasing sea-ice cover reconstructed through most of the Late Holocene is presumed to favor glacier growth (suppressing summer temperatures, minimizing the number of positive degree days and decreasing frontal ablation) it is unclear how much this restricts precipitation.
Another indicator for Holocene environmental conditions on Svalbard is the formation of permafrost landforms and ground ice (Humlum et al., 2003;Humlum, 2005).A sedimentological investigation detailing the formation of permafrost in lowland terrain from central Spitsbergen highlights the inter-connection of relative sea level and permafrost aggradation during the Late Holocene (Gilbert et al., 2018).This work compliments earlier studies which present maximum constraining ages from the last 4.0 ka BP of various permafrost landforms and periglacial deposits, including pingos (Yoshikawa and Nakamura, 1996), ice-wedge polygons (Jeppesen, 2001) and aeolian loess deposits (Oliva et al., 2014).Generally, lowland permafrost landforms display a close correlation with uplift emergence patterns and Neoglacial cooling (Yoshikawa and Nakamura, 1996;Gilbert et al., 2018).Ground ice however, is also identified above the post-glacial marine limit, and potentially formed earlier in the Holocene (or in some cases earlier in the Quaternary) although ice-bodies generally lack the same detail of chronological constraint as investigated lowland ground ice (Humlum, 2005;Humlum et al., 2007;Siewert et al., 2012;Eckerstorfer et al., 2013).

Glacier response to Holocene cooling events
Holocene "cooling events" on Svalbard associated with meltwater pulses or cold episodes within the North Atlantic (Bond et al., 2001;Jennings et al., 2015) have been proposed by various proxies from lake and fjord records, suggesting distinct events e.g.11.3 ka BP (Tian et al., 2020), 9.3 ka BP (van der Bilt et al., 2019), 8.2 ka BP (Hald andKorsum, 2008) and4.2 ka BP (van der Bilt et al., 2016).Despite proxy records promoting distinct environmental shifts during these periods in the Holocene, no glacial landform on Svalbard has been constrained to any of these events.In comparison, ice-marginal reconstructions from both sides of Baffin Bay seem well behaved, exhibiting synchronous breaths of Early -Middle Holocene punctuated decline (Lesnek and Briner, 2018;Young et al., 2020).However, there are no indications of a similar synchronous glacier response in Svalbard.Consequently, these cooling events, although some cases influential in Svalbard's Holocene environmental reconstructions, were not effective drivers of Svalbardwide Holocene glacier activity.This may indicate that the extent of glacier mass loss experienced during the Early -Middle Holocene forced cooling-event-ice-margins inside of what would later become the Late Holocene maximum extent.Furthermore, although periods of reduced summer temperatures may halt ablation, they do not (necessarily) recharge Svalbard's glacier catchments.Subsequently, the interconnection between cool and dry versus warm and wet may-be critical with respect to the slow glacier turn-over characteristic of Svalbard's High Arctic glaciers (Wanner et al., 2001;Østby et al., 2017).
In comparison to these relatively short cooling events, the extended and episodic phase of Late Holocene cooling forced by declining high latitude summer insolation and explosive volcanism (Laskar et al., 2004;Miller et al., 2017;Bradley and Bakke, 2019), effectively revitalized Svalbard's modern glaciers (Bakke et al., 2018).Late Holocene glacier re-advances have been identified in terrestrial and marine archives as well as constrained from glacial lake and fjord sedimentary records (Fig. 17).Throughout the Late Holocene the percentage of inferred ice-cover on Svalbard may have ranged from as low ~25 %, up to nearly 75 % during the Late Holocene glacier maximum (LHM), while regional snowline lowered and reconstructed equilibrium line altitude (ELA) dropped ~250 m (Fig. 17; Røthe et al., 2015;Miller et al., 2017).
Environmental reconstructions of lakes and terrestrial conditions (e.g.vegetation and permafrost ground-ice) suggest temperatures progressively cooled over the last 5 ka (Fig. 17).Marine records also suggest fjord temperatures episodically declined in step with sea-ice cover transitioning from marginal to extended (Fig. 17).Ice-margins extend in episodic fashion through the Late Holocene, culminating in a period with peak ice-cover and extended glacier-systems (Fig. 17 Werner, 1993;Svendsen and Mangerud, 1997;Miller et al., 2017;Philipps et al., 2017).While the Neoglacial exhibited no marked event signaling its initiation, persistent climatic deterioration through the second half of the Holocene re-invigorated Svalbard's glaciers (Bradley and Bakke, 2019).

2015).
While Early Holocene glacier re-advances may not draw much attention, being mere interruptions in deglaciation, the process understanding has the potential to improve projections of ongoing and future ice loss.Whether Early Holocene glacier re-advances were driven by internal dynamics or mass-balance, they were short-lived and unsustainable glacier fluctuations.By kinematicly passing long accumulated ice mass below an ELA and spreading it within an ablation zone, the (dynamic) rate of ice loss will exceed non-dynamic ablation (Willis et al., 2018).Thus, a glacier advancing into a warming climate will undergo greater mass loss than a glacier-system at equilibrium in the same climate.
Early Holocene glacier re-advances undoubtedly increased a small percentage of ice cover on Svalbard at various points, yet (unlike the Neoglacial) were not necessarily a result of positive glacier mass balance (Fig. 8, 13 & 17).Reconstructed Early Holocene warmth derived from fjord, lake and terrestrial records suggests climate conditions were not entirely favorable for glacier mass accumulation (Fig. 17).Consequently, the key driver of Early Holocene glacier activity is unknown.Constraints on the amount Early Holocene precipitation are need to evaluate if regions experienced periods of positive mass balance, effectively outweighing glacier retreat (Thomas et al., 2018).Additionally, enhanced precipitation may have been factored by the severity of Early Holocene isostatic uplift resulting in an environment where glaciers traveled up through their ELAs (Norðdahl and Ingólfsson, 2015).Conceptually, glacier ablation area would have decreased while accumulation zones extended (Farnsworth, 2018).These processes may have driven Early Holocene glacier fluctuations, yet remain to be tested and modeled.
Observations during 20 th century warming has led to developments in process understanding which may serve as analogues to Early Holocene conditions.For example, modern enhanced precipitation has been correlated to warming conditions and decreasing sea-ice (Nowak and Hodson, 2013;Bintanja and Selten, 2014), yet glacier accumulation on Svalbard has not outpaced ablation (Østby et al., 2017).However, still some glacier-systems has continued to exhibit ice-marginal advances (Dunse et al., 2015;Lovell et al., 2018;Sevestre et al., 2018).Ice dynamics have played an important role in 20 th century glacier fluctuations (Flink et al., 2015;Sevestre and Benn, 2015;Farnsworth et al., 2016;Lovell et al., 2018), yet the extent of ice dynamics driving icemarginal fluctuations at different periods (either warm or cold) in the Holocene in unknown.
In the surge-type glacier enthalpy model introduced by Sevestre and Benn (2015), a theoretical climatic envelope is constructed around a global inventory of glaciers, suggesting an ideal environment for glacier surge behavior.Given shifting climatic conditions, it is possible to have a glacier-system move from outside to inside this theoretical climatic envelope.The enthalpy model aligns with a conceptual model used to explain the deposition of transitional landforms during the Late Glacial deglaciation on the West coast of Svalbard (Landvik et al., 2013(Landvik et al., , 2014)).The conceptual model suggests a shift in glacier basal conditions was a likely contributor to ice marginal re-advances during the deglaciation (Landvik et al., 2013).Interestingly, recent developments in process understanding of modern surge behavior on Svalbard suggests that warming environmental conditions has effected glacier thermal regime, hydrologic system and dynamics which has had a direct effect on glacier basal conditions and ice-marginal extent (Dunse et al., 2015;Sevestre et al., 2018).While these surge events have increased the percentage of ice-cover on Svalbard, they have not been fostered from phases of positive mass balance.The change in percentage ice-cover has been a result of a shift in basal conditions driving the kinematic redistribution of mass and a dynamic advance as the profile of the glaciersystem flattens (Hagen, 1987;Sevestre et al., 2018;Willis et al., 2018).
While we have developed a general understanding of shifts in Svalbard's ice cover throughout the Holocene, our understanding of shifts in glacier basal conditions remain limited, yet may play a critical role in Holocene glacier marginal fluctuations.The Early Holocene deglaciation was punctuated by glacier re-advances during what is reconstructed to be an exceptionally warm environment.These re-advances appear to relate to the time-transgressive nature of deglaciation rather than a glacier event driven by a short period of cooling or an interval of enhanced precipitation yet this remains to be tested.Regardless of driving factor, glacier re-advances during a warm event reflect the complex style and efficient dynamic mass loss of glaciers in a changing climate.

Summary and perspective
Throughout the Holocene, Svalbard glaciers have responded to a varying combination of climatic, environmental and dynamic driving factors which influence both the extent and behavior of ice margins.Holocene glacier history, dynamics and processes provide key insights into glacier response to shifting climate and environment.Our understanding of Svalbard's ice-marginal fluctuations and glacier behavior through the Holocene allows us to refine our projections of future glacier conditions.− Ice cover on Svalbard during the transition from Late Glacial to Early Holocene was likely less extensive than previously reconstructed (e.g.DATED-1).However, the region was still more extensively glaciated than any other period during the interglacial.Future investigations focusing on this transition should strive to further define ice-margins and test whether deglacial ice-shelves had an effect on ice-marginal landform development or glacier dynamics.− Svalbard fjord systems deglaciated rapidly, in a time-transgressive fashion, during the first half of the Early Holocene (12.0 -10.5 ka BP) which environmental reconstructions suggest exhibited exceptional regional warmth.While fjord system retreat-percentage provides a perspective on the timing of deglaciation, retreat rates provide a simplistic range of potential scenarios of ice loss.− Early Holocene ice retreat on Svalbard was punctuated by widespread glacier re-advances, presumably dynamic and unsustainable (i.e., not driven by positive mass balance).Glacier behavior served as an efficient mechanism of deglaciation and reflects the complex style of ice-mass-loss during a changing climate.Modelling environmental conditions capable of driving Early Holocene glacier fluctuations may elucidate key driving factor(s) behind this icemarginal behavior.− The Holocene glacial minimum on Svalbard occurred during the beginning of the Middle Holocene (between 8.0 -6.0 ka BP).However, a small portion of Svalbard glaciers (~25 % land-cover) located in the north and east of Svalbard presumably survived the minimum with some fjord systems hosting calving margins throughout the entire interglacial.In order to test these findings, evidence should be sought from other archives (lacustrine or ice core records) which indicate glaciers persistent through the glacial minimum.− No glacial landforms have been identified corresponding to glacier re-advances or still-stands (9.0 -4.5 ka BP) during the Middle Holocene.Consequently, relatively little is known about the extent of Svalbard glaciers during the Middle Holocene other than lake and some fjord records suggesting catchments hosted severely reduced ice cover.Better understanding glacial minimum conditions would improve (glacier, climate and isostatic adjustment) model domains and provide a potential perspective of Svalbard's future glaciers.− Late Holocene glacier re-advances and corresponding deposits have been attributed to episodic Neoglacial cooling and the LIA.Glacier re-advances have predominantly been dated to between 4.0 -0.5 ka BP with the highest frequency constrained between 1.0 -0.5 ka BP.Much like Early Holocene glacier activity, the role of precipitation and seasonality versus summer temperatures remains poorly defined.Past precipitation patterns likely exhibited significant variability and impact on Holocene glaciers.Any development will enhance a critical (yet poorly constrained) component of the mass balance of Svalbard's glaciers.− Despite post-LIA glaciers on Svalbard exhibiting widespread negative mass balance, ice marginal retreat, and glacier thinning, some Svalbard glaciers have continued to exhibit surge-type re-advances.Dynamic glacier behavior has been linked to shifts in thermal regime, hydrologic system and surface profile, likely reflecting a similar pattern as the dynamic deglaciation during the Early Holocene.Ultimately, an enhanced understanding of Holocene glaciers and their driving factors will foster better projections of glaciers in response to changing climate.− Syntheses of existing geochronological data are useful for identifying gaps in data and knowledge.While Svalbard currently exhibits relatively widespread geochronological constraint of glaciers and environment through the Holocene, holistic multidisciplinary investigations could be developed to target known-unknowns related to glaciers, geochronology and environment.One venue that has the potential to enhanced geochronology constraint in all (and in between) records, is further detailing the distal deposition of tephra on Svalbard.Defining a Holocene chronostratigraphic tephra framework for Svalbard, based on pan-Arctic volcanism, would strengthen high-Arctic lacustrine terrestrial and marine chronologies.Additionally, a framework of (crypto-)tephra horizons would allow for time-synchronous comparison between archives with the potential to further resolve leads and lags between terrestrial, lacustrine and marine systems.
(UNIS) (research grant to ÓI), and Arctic Research and Studies funded by the Ministries for Foreign Affairs of Norway and Iceland (2017-ARS-79772 to AS & WRF).We are grateful for extensive support from the UNIS Library (Berit Jakobsen and Catherine Fjeldstad).Constructive comments from and discussions with Skafti Brynjólfsson, Kelly Hogan, Anna L.C. Hughes, Sofia Kjellman, Nicolaj K. Larsen and Anne Sofie Søndergaard have improved this manuscript.We are also thankful for anonymous reviews as well as constructive comments from Managing Editor, Professor Negri.

Fig. 1 .
Fig. 1.Location maps with A) inset map of the polar North Atlantic region with the Svalbard archipelago framed in black box.B) Overview map of Svalbard identifying key islands and regions.The warm West Spitsbergen Current (WSC) flows northward along Svalbard′s western margin while the cool East Spitsbergen Current (ESC) surrounds the southern tip of Spitsbergen from the east.Figure maps modified from IBCAO (Jakobsson et al., 2012) and TopoSvalbard © Norwegian Polar Institute 2020 respectively.

Fig. 2 .
Fig. 2. Modified photograph taken of Trygghamna in 1908 by Oscar Halldin.Glaciers Protektorbreen and Harrietbreen with ice-margins calving into the bay beyond Swedish expedition ships.Photograph restoration and editing by Erik S. Holmlund.Expeditions described by De Geer (1908).

Fig. 3 .
Fig. 3. Holocene landforms and sedimentary archives have been used to provide insight into the glacier history of Svalbard, here with examples from De Geerbukta, NE Spitsbergen.Aerial images from TopoSvalbard © Norwegian Polar Institute 2020 and fjord bathymetry modified from Streuff et al. (2017).

Fig. 4 .
Fig. 4. Spatial distribution of ages compiled in the SVALHOLA database.Colored circles are proportional to the number of dates of each method per site.Note the low density of chronological data from eastern Svalbard and the east coast of Spitsbergen.The bulk of the ages are radiocarbon dates from low elevation coastal regions and from shallow waters to the west of Spitsbergen.Dates were compiled from references presented in TableS1-S3 and at the end of this manuscript (full references are included in the supporting information; Data S1).

Fig. 7 .
Fig. 7. Transects highlight Early Holocene ice retreat within fjord systems; Bellsund, Isfjorden, St. Jonsfjorden, Woodfjorden, Wijdefjorden and Wahlenbergfjorden.Plots present retreat length (km) and percentage retreat (%) over time separately for each fjord system and comparatively for all fjord systems.Fjord length calculations are based on the distance between database ages rather than modern fjord dimensions.Thus, fjord lengths are in some cases underrepresented as ages do not reach modern fjord heads while other fjord lengths are overrepresented as ages are found in fjord valleys, presently located kilometers inland.

Fig. 8 .
Fig.8.Distribution of Early Holocene glacier re-advances identified around Svalbard.Re-advances occur from 12 -9 ka BP with greater than half of the glacier readvances happening between 11 -10 ka BP.Dates were compiled from references presented in TablesS1-S3and at the end of this manuscript (full references are included in the supporting information; Data S1).

Fig. 9 .
Fig. 9. Diagram highlighting the chronology and distribution of glacial lake records on Svalbard used to reconstruct Holocene glacier activity.Blue bars indicate high (dark blue) and low/reduced (light blue) minerogentic input within the lake records.Minerogentic input has be used to suggest the presence or absence of an active glacier within the lake catchment.Terrestrial plant macrofossils (black boxes), crypto-tephra (red boxes), marine mollusc shells and foraminifera (blue boxes) and bulk sediment ages (white boxes) are used to constrain core chronologies.Values to the left of the chronologies indicate the number of dates and the composite core length of each record.Inset map shows the location of the different records: G = Gjøavatnet (de Wet et al., 2018), V = Vårfluesjøen(Mäusbacher et al., 2002;Røthe et al., 2018), H = Hajeren (van der Bilt et al., 2016), K = Kløsa(Røthe et al., 2015), Kb = Kløverbladvatnet(Schomacker et al., 2019) and L = Linnévatnet(Svendsen and Mangerud, 1997;Svendsen and Mangerud, 1997;Snyder et al., 2000).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 10 .
Fig. 10.Diagram highlighting the age and distribution of investigated glacial influenced fjord systems used to reconstruct Holocene glacier activity.Blue bars indicate high (dark blue) and low/reduced (light blue) ice rafted debris within the marine records.Ice-rafted debris concentrations have been used to suggest the presence or absence of a glacier calving margin within the fjord system.Terrestrial plant macrofossils (black boxes), marine mollusc shells and foraminifera (blue boxes) are used to constrain core chronologies.Values at left above chronologies indicate the number of dates and the composite core length of each record.Inset map shows the location of the different records.Fjords: Wf = Wahlenbergfjorden ( 1 Flink et al., 2017; 2Bartels et al., 2018), Wdf = Woodfjorden(Bartels et al., 2017), Kf = Kongsfjorden(Skirbekk et al., 2010), Bf = Billefjorden(Baeten et al., 2010), Tf = Tempelfjorden(Forwick et al., 2010), VMf = Van Mijenfjorden(Hald et al., 2004), VKf = Van Kuelenfjorden(Kempf et al., 2013), Sf = Storfjorden(Nielsen and Rasmussen, 2018).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 4.3.Middle Holocene (8.2 -4.2 ka BP)

Fig. 16 .
Fig.16.Plot of 497 terrestrial radiocarbon ages and raised marine from the SVALHOLA database from the last 6 ka BP.Grey column delimits the end of the Middle Holocene (4.2 ka).Specific radiocarbon ages from the database have been highlighted as colored circles with error bars based on palaeoglaciological and climatological associations.Histogram at base indicates TCN exposure age distribution within 500 year bins for moraine ridge boulders three sites, L = Linnébreen, S = Scottbreen and T = Treskelen(Reusche et al., 2014;Philipps et al., 2017).Shaded bars at top represent a schematic view of Late Holocene glacial activity reconstructed from lake sediments of four proglacial lakes from western Spitsbergen (Gjøavatnet, Hajeren, Kløsa and Linnévatnet).Blue boxes represent periods of glacial activity (dark blue suggest enhanced glacial activity), red boxes suggest no glacier was present in the catchment or glacial activity was greatly reduced; gray boxes indicate periods of reduced organic matter accumulation (Gjøavatnet; modified from deWet et al., 2018).Light blue boxes indicate periods of glacier growth interrupted by moraine stabilization (white boxes) reconstructed with lichenometry(Werner, 1993).Data available in TablesS1 & S2.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) ;