Holocene sea-ice dynamics in Petermann Fjord in relation to ice tongue stability and Nares Strait ice arch formation

. The Petermann 2015 Expedition to Petermann Fjord and adjacent Hall Basin recovered a transect of cores from 15 Nares Strait to under the 48 km long ice tongue of Petermann glacier, offering a unique opportunity to study ice-ocean-sea ice interactions at the interface of these realms. First results suggest that no ice tongue existed in Petermann Fjord for large parts of the Holocene, raising the question of the role of the ocean and the marine cryosphere in the collapse and re-establishment of the ice tongue. Here we use a multi-proxy approach (sea-ice related biomarkers, total organic carbon and its carbon isotopic composition, and benthic and planktonic foraminiferal abundances) to explore Holocene sea-ice dynamics at OD1507-03TC- 20 41GC-03PC in outer Petermann Fjord. Our results are in line with a tight coupling of the marine and terrestrial cryosphere in this region


Introduction
Nares Strait, connecting the Lincoln Sea to the northern Baffin Bay, is an important conduit for sea ice, freshwater and heat 35 between the Arctic Ocean and the western North Atlantic. The annual flux of freshwater through Nares Strait, in liquid and solid form, heavily depends on the seasonal formation of ice arches (Kwok et al., 2010;Münchow, 2016;Rasmussen et al., 2010). Ice arches form when drift ice converges in a narrow passage between two landmasses. In Nares Strait their formation depends primarily on the sea-ice thickness, local wind stresses, and atmospheric temperatures (Barber et al., 2001;Kwok et al., 2010;Samelson et al., 2006). Ice arching inhibits sea-ice export from the Arctic Ocean and allows the formation of landfast 40 ice in Nares Strait, consisting of a mixture of multi-year drift ice, originating from the Arctic Ocean, and locally formed firstyear ice (Kwok, 2005;Kwok et al., 2010). Historically, the formation of a northern and southern arch have been observed in Robeson Channel and Smith Sound, respectively ( Fig. 1) (Vincent, 2019). In recent decades, however, changes in the ice arch configuration suggest a transition in Nares Strait sea-ice dynamics. Between 1979 and 2019, Nares Strait was blocked for seaice passage on average 161 days per season with a consistent decrease of 2.1 days/year throughout this period (Vincent, 2019). 45 This is also associated with an emerging prominence of the northern arch (Vincent, 2019). In the winter of 2006/2007 both ice arches failed to form for the first time in recorded history, causing sea ice to remain mobile in Nares Strait year-round (Kwok et al., 2010;Vincent, 2019) (Fig. 2, Supplementary Fig. 1). The observed changes in Nares Strait sea-ice dynamics likely have significant consequences for the export of multi-year sea ice from the Lincoln Sea and long-term Arctic sea-ice loss (Kwok et al., 2010;Moore et al., 2021;Vincent, 2019) (Fig. 2). Additionally, the formation of the southern ice arch in Smith Sound is 50 crucial for the annual opening of the North Water Polynya (NOW) (Fig. 1) and the formation of landfast sea ice in Nares Strait (Barber et al., 2001). The latter has important implications for the hydrographic structure in Nares Strait and its adjacent fjords in response to changing wind stresses on surface waters in Nares Strait (Shroyer et al., 2015(Shroyer et al., , 2017. The water column in Nares Strait is characterized by cold and fresh Polar Water (PW) in the upper 50-100 m with warmer and more saline modified/Arctic Atlantic Water (AAW) below, separated by a strong halocline (Johnson et al., 2011;Münchow et 55 al., 2014). Under landfast sea ice, Ekman transport causes eastward displacement of cold and fresh PW towards the Greenland coast (Rabe et al., 2012;Shroyer et al., 2015Shroyer et al., , 2017. Conversely, mobile sea ice leads to westward Ekman transport of PW and upwelling of AAW in the east, increasing the oceanic heat flux to fjord systems along the Greenland coast of Nares Strait (Münchow et al., 2007;Shroyer et al., 2017). Outlet glaciers draining into fjords in the north and northeast of Greenland, commonly terminate in a floating ice tongue of variable length, with three glaciers terminating in an ice tongue >10 km (Hill 60 et al., 2017). One such glacier is Petermann Glacier (PG), draining about 4 % of the Greenland Ice Sheet (GrIS) (based on area of the drainage basin (Rignot and Kanagaratnam, 2006)) into Hall Basin in Nares Strait (Fig. 1). Large calving events of PGs floating ice tongue in 2010 and 2012 (Johannessen et al., 2013;Rückamp et al., 2019) were associated with a 10 % acceleration of the glacier (Rückamp et al., 2019). Interestingly, the 2010 calving event occurred at the end of a 4-year period with no/little landfast ice in Nares Strait and associated earlier break-up of landfast ice in Petermann Fjord (Fig. 2). The smaller calving 65 Submarine melt rates depend on the turbulent heat flux reaching the ice tongue/ocean interface. This is a function of the oceanic heat content, determined by the inflow of AAW to Petermann Fjord, and turbulent mixing underneath the ice tongue, promoted by subglacial meltwater discharge (Cai et al., 2017;Washam et al., 2018). Thus, changes in Nares Strait sea-ice dynamics, that modify AAW inflow to Petermann Fjord, may affect the ice tongue stability on long timescales. Additionally, earlier work has 90 identified landfast sea ice in fjords around Greenland as an important mechanism stabilizing the calving front of marineterminating outlet glaciers, with the loss of landfast ice prolonging the calving season (Amundson et al., 2010;Carr et al., 2015;Robel, 2017;Todd and Christoffersen, 2014). Hence, in light of the recent doubling of mass loss from the GrIS (Shepherd et al., 2012) a detailed understanding of ocean-sea ice-glacier interactions on longer timescales is essential to improve projections for the contribution of the GrIS to future sea level rise (Fürst et al., 2015;Stocker et al., 2013). 95 In 2015, the Swedish Icebreaker Oden set out for The Petermann 2015 Expedition to improve our understanding of the processes involved in ocean-sea ice-glacier interactions and the sensitivity of PGs floating ice tongue to Holocene climate change. A transect of sediment cores, extending from Hall Basin to underneath the Petermann ice tongue, was recovered (Reilly et al., 2019). Spatial differences in sediment facies associated with the presence/absence of the ice tongue allowed the reconstruction of the extent of the Petermann ice tongue over the last ~7,000 cal yrs BP (Reilly et al., 2019). Reilly et al. (2019) 100 demonstrate that after the deglacial break-up of the ice tongue at ~6,900 cal yrs BP, no stable ice tongue existed in Petermann Fjord for the largest part of the mid-Holocene, with a small ice tongue re-emerging at ~2,200 cal yrs BP, which advanced to its modern limits around 600 years ago.
Here, we focus on the Holocene evolution of sea-ice conditions in Petermann Fjord. The spliced sediment core OD1507-03TC- .023° E; 976 m water depth) located in outer Petermann Fjord (Fig. 1), offers a unique opportunity 105 to study Nares Strait and local sea-ice dynamics, and their influence on the stability of PG. Sea-ice reconstructions are based on source-specific Arctic sea-ice biomarkers (Belt, 2018). Measurements of total organic carbon (TOC), the carbon isotopic composition of TOC, sterol biomarkers and the benthic and planktonic foraminiferal abundance provide information with regard to marine primary productivity and terrestrial organic carbon input to Petermann Fjord across the Holocene. In combination with existing studies (e.g. England et al., 2008;Funder et al., 2011;Georgiadis et al., 2020) our results offer 110 insights into the Holocene development of ice arches in Nares Strait. Importantly, this study demonstrates that the development of more severe sea-ice conditions in Petermann Fjord preceded major advances of the ice tongue, indicating a stabilizing effect on PG.

Regional oceanography
At the northern end of Nares Strait, Robeson Channel connects Hall Basin to the Lincoln Sea ( Fig. 1) with water mass exchange 115 controlled by a 290 m deep sill (Münchow et al., 2011a;Washam et al., 2018). In the upper 50 m (off Greenland) to 100 m (off Ellesmere Island) (Jones and Eert, 2004;Münchow et al., 2007), Nares Strait is characterized by relatively fresh and nutrient-rich Polar Water (PW). Geochemical tracers suggest that these waters are primarily of North Pacific origin, entering the Arctic Ocean via the Bering Strait, modified by river runoff and sea-ice melt (Jones and Eert, 2004;Münchow et al., 2007).
Below (>300 m in Hall Basin), the water column is characterized by AAW (0.28-0.31 °C; (Washam et al., 2018)) (Jones and 120 Eert, 2004). Hydrographic surveys have shown that the AAW in Nares Strait has warmed by 0.023 ± 0.015 °C per year between (Münchow et al., 2011b. At its shallowest, at the northern end of Kane Basin, Nares Strait is 220 m deep (Münchow and Melling, 2008). This sill impedes the southward flow of AAW, suggesting that Atlantic Water at the southern end of Nares Strait is predominantly derived from the north flowing West Greenland Current (Fig. 1) (Melling et al., 2001).
The circulation in Nares Strait is dominated by a southward surface jet controlled by winds and along channel pressure 125 gradients between the Lincoln Sea and northern Baffin Bay (Rabe et al., 2010(Rabe et al., , 2012. The hydrographic structure in Nares Strait varies according to the predominant sea-ice state (Rabe et al., 2012;Shroyer et al., 2015Shroyer et al., , 2017. Modelling studies show that this is a response to surface stresses (Shroyer et al., 2015(Shroyer et al., , 2017. Landfast sea ice exerts a northward drag at the ocean surface, resulting in eastward Ekman transport of cool and fresh PW and a westward shift of the main surface jet towards Ellesmere Island (Rabe et al., 2012;Shroyer et al., 2015Shroyer et al., , 2017. During the mobile sea-ice season southward wind stress and 130 associated westward Ekman surface transport cause a displacement of cool and fresh surface waters towards Ellesmere Island and upwelling of relatively warm and salty waters along the Greenland coast, while the main southward flow is concentrated in the centre of the channel (Münchow et al., 2007;Shroyer et al., 2017). Nares Strait is covered by sea ice for around 10-11 months per year, with a 95% ice cover during winter months (Rasmussen et al., 2010). Summer break-up occurs in June/July, with renewed freeze up in late September/October. The flow of sea ice through Nares Strait is highest during fall and early 135 winter (Kwok et al., 2010) with large interannual variability depending on the formation and duration of the northern and southern ice arch. While the formation of either arch will block the export of Arctic sea ice through Nares Strait, only the formation of the southern arch leads to the opening of the NOW (Barber et al., 2001) and a complete freeze up of Nares Strait (Kwok, 2005;Kwok et al., 2010).
Petermann Fjord, connected to Nares Strait (Hall Basin) via a 350-450 m deep sill, is up to 1100 m deep and ~20 km wide 140 (Jakobsson et al., 2018). It hosts a floating ice tongue, approximately 48 km long (from the grounding zone) with an average width of 16.6 km and a thickness of 600 m at the grounding zone to 200 m at the terminus (Heuzé et al., 2017;Johannessen et al., 2013). The ice tongue flows at a speed of 1250 ± 90 m yr -1 over the grounding zone, resulting in a calculated net glacial freshwater flux of 0.26 mSv (Heuzé et al., 2017). The formation of landfast ice in Petermann Fjord is somewhat independent of the formation of landfast ice in Nares Strait. Landfast ice in Petermann Fjord will also form when sea ice in Nares Strait 145 remains mobile throughout the winter (e.g. 2007, 2009, 2010Fig. 2), although the sea-ice state in Nares Strait likely influences the timing of sea-ice break-up in Petermann Fjord in spring/summer, with earlier/later break-up during years without/with landfast sea ice in Nares Strait (Fig. 2, Supplementary Fig. 1). Additionally, we propose that the sea-ice dynamics in Nares Strait likely have important implications for the primary productivity regime in Petermann Fjord, with ice edge conditions in spring/ early summer in years with mobile sea ice in Nares Strait (e.g. 2007 and2009;Supplementary Fig. 1), while under 150 landfast ice conditions in Nares Strait the spring/early summer ice edge is situated several hundred kilometres to the southwest in Kane Basin/Smith Sound (e.g. 2013 and 2014; Supplementary Fig. 1). Since the spring ice edge is a highly productive system, especially important for the sympagic algal bloom (Ardyna and Arrigo, 2020;Leu et al., 2015;Wassmann and Reigstad, 2011), the position of the Nares Strait ice edge is likely to play a role for the local primary productivity. The hydrographic structure in Petermann Fjord is closely linked to that in Nares Strait with colder, fresher PW overlaying AAW (Johnson et al., 2011;Münchow et al., 2014). Bottom waters in Petermann Fjord are renewed by episodic spillover of AAW from Hall Basin, with bottom water properties in Petermann Fjord resembling those at ~380 m water depth in Hall Basin (effective sill depth) (Johnson et al., 2011). In general, the circulation in Petermann Fjord resembles an estuarine model, with 170 outflow of buoyant, meltwater-enriched surface waters along the northeast side of the fjord and inflow of AAW below, concentrated along the south-western side of the fjord mouth (Heuzé et al., 2017;Johnson et al., 2011;Washam et al., 2018).
In the fjord mouth, eddy structures can enhance the exchange between Hall Basin and Petermann Fjord. Eddies are stronger and more stable during summer, when sea ice in Nares Strait is mobile (Johnson et al., 2011;Shroyer et al., 2017). Modelling studies suggest that the displacement of water masses in Nares Strait in response to the prevailing sea-ice regime also affects 175 Petermann Fjord, with enhanced inflow of warmer, saltier AAW during times of mobile sea-ice (Shroyer et al., 2017).
Additionally, a stronger circulation in the fjord, driven by enhanced subglacial runoff during summer months, increases the transport of AAW to the ice tongue cavity and the turbulent mixing of AAW toward the base of the ice tongue (Cai et al., 2017;Washam et al., 2018). In response to warming of AAW in Nares Strait (Münchow et al., 2011b), a 0.2°C warming of AAW in Petermann Fjord has been observed from 2002 to 2016 (Washam et al., 2018). In combination, the greater oceanic 180 heat flux and strengthened under ice currents cause enhanced submarine melting and non-steady state thinning of the ice tongue (Cai et al., 2017;Washam et al., 2018Washam et al., , 2019.

Sediment core OD1507-03TC-41GC-03PC
Sediment core OD1507-03TC-41GC-03PC is stored at the Oregon State University Marine and Geology Repository at 2.8°C. 185 Samples were taken approximately every 5 cm for the analysis of different organic constituents and approximately every 10 cm for benthic and planktonic foraminiferal analyses. The core represents a spliced record of a trigger core (TC), a gravity core (GC), and a piston core (PC) recovered in outer Petermann Fjord as part of The Petermann 2015 Expedition ( Fig. 1) ( Table 1). The splice has a total length of 556 cm and was recovered 80 km from the 2015 PG grounding zone at an average water depth of 976 m (Reilly et al., 2019). Based on Computed Tomography (CT) scans, continuous sections with minimal 190 disturbance were chosen for the splice and correlation was performed using X-Ray Fluorescence (XRF) Ti/Ca ratios, CT slice images, and CT numbers (Reilly et al., 2019) (Supplementary Table 1). Three sedimentary units are distinguished in OD1506-03TC-41GC-03PC, based on physical properties and XRF elemental data (Reilly et al., 2019). Sedimentary unit 3 (ca. 518-555 cm) is a massive diamict composed of a sandy mud with abundant coarse clasts and XRF Ti/Ca ratios around 0.05 (Reilly et al., 2019). The clasts found in this unit, likely do not represent ice rafted debris (IRD), but instead are related to proximity 195 of the grounding zone (Reilly et al., 2019). Unit 2, a clayey laminated mud with no or very low concentrations of coarser material (IRD), is found between ca. 398-518 cm (Reilly et al., 2019). The topmost lithofacies, unit 1 (0-398 cm), is a bioturbated clayey mud with isolated sand and coarser particles (Reilly et al., 2019). Based on the XRF Ti/Ca ratios and the abundance of coarse material, unit 1 has been divided into three subsections (A (0-53 cm), B (53-164 cm), C (164-398 cm)) (Reilly et al., 2019). Subunit 1C is marked by a decreasing trend from high to intermediate IRD occurrence, with intermediate 200 IRD abundances continuing during subunit 1B, followed by low IRD abundance during subunit 1A (Reilly et al., 2019).
The age model for OD1507-03TC-41GC-03PC, established by Reilly et al. (2019), is based on radiocarbon ( 14 C) dating of benthic and planktonic foraminifera and calibration of radiocarbon ages using the Marine13 curve (Reimer et al., 2013) and MatCal MATLAB tools (Lougheed and Obrochta, 2016). The ΔR value was constrained using Paleosecular Variation (PSV) stratigraphy, with the best fit determined for a constant ΔR choice of 770 years (Reilly et al., 2019). No ages were determined 205 for the interval between 408 cm and 556 cm, due to lack of radiocarbon dates. Regional constraints, however, suggest that these sediments are younger than 7,600 years, corresponding to the timing of the retreat of PG into the fjord (Jakobsson et al., 2018).

Sea-ice biomarker methodology 210
Reconstructions of past sea-ice conditions in Petermann Fjord rely on the identification of source-specific Arctic sea-ice biomarkers, together with the identification of common sterol biomarkers, TOC, and planktonic and benthic foraminiferal abundances. This is a qualitative method to determine past sea-ice dynamics. For clarity, we will describe past sea-ice conditions using the following categories in order of increasing average sea-ice concentration: ice free, reduced seasonal sea ice, enhanced seasonal sea ice, and near-perennial. 215 Source-specific sea-ice biomarkers include a mono-and a di-unsaturated highly branched isoprenoid (HBI), termed IP25 (ice proxy with 25 carbon atoms) and HBI II (Belt, 2018;Belt et al., 2007). IP25 is produced by a number of spring sea-ice dwelling diatoms, including Haslea spicula, H. kjellmanii, and Pleurosigma stuxbergii var. rhomboides (Brown et al., 2014). Thus, its presence in Arctic marine sediments provides evidence for past seasonal sea-ice occurrence, while the absence of IP25 occurs in year-round ice-free environments as well as under perennial sea-ice cover (Belt, 2018;Belt et al., 2007;Brown et al., 2014;220 Navarro-Rodriguez et al., 2013;Xiao et al., 2015). Salinity changes might exert an additional control on IP25 production, with Ribeiro et al. (2017), showing that its production appears to be suppressed by meltwater in fjords of North Eastern Greenland.
Given its co-production in H. spicula, H. kjellmanii, and Pleurosigma stuxbergii var. rhomboides (Brown et al., 2014), HBI II co-varies with IP25 in the Arctic realm. Thus, the typically higher HBI II concentrations can provide additional information during times of low/absent sedimentary IP25. 225 In recent years, biomarker-based reconstructions of past sea-ice dynamics have commonly also reported a tri-unsaturated HBI (HBI III), produced by diatoms characteristic of the spring sea-ice edge bloom in the marginal ice zone (MIZ) (Belt et al., 2015(Belt et al., , 2017Smik et al., 2016). In Petermann Fjord, however, the identification of the HBI III peak in gas chromatography-mass spectrometry (GC-MS) chromatograms was compromised due to interference with a neighbouring peak ( Supplementary   Fig. 2). This is specific to sediment samples from Petermann Fjord, as HBI III could be identified in reference sediment from 230 Young Sound extracted alongside the samples from Petermann Fjord (see 3.3.2). Instead, we use a range of sterol biomarkers (campesterol, cholesterol, brassicasterol, dinosterol, and β-sitosterol) and the abundance of benthic and planktonic foraminifera to gain insights into the regional primary productivity regime and the addition of terrestrial organic carbon to Petermann Fjord.
Sterols are common compounds in eukaryotic cell membranes, abundant in both marine and terrestrial organic carbon. This complicates their use as unambiguous tracers of a specific organism and/or environmental regime (Belt and Müller, 2013;235 Rontani et al., 2014;Volkman, 1986). Nonetheless, the relative abundance of certain sterols can vary according to the predominant organic carbon source (Rontani et al., 2014;Volkman, 1986;Yunker et al., 1995) (Table 2).
Here we use brassicasterol, dinosterol, and cholesterol to represent phytoplankton primary productivity in Petermann Fjord, with different relative concentrations, potentially indicating changes in the ecosystem composition. Campesterol and β-245 sitosterol, on the other hand, are used to gain insight into terrestrial organic carbon input. The multiproxy study of IP25, HBI II, sterol biomarkers, and the benthic and planktonic foraminiferal abundance (see section 3.4), thus allows interpretations of past sea-ice dynamics and terrestrial versus marine organic carbon input to Petermann Fjord.

Analysis of total organic carbon (TOC), carbon isotopic composition of TOC (δ 13 Corg), and sea-ice related biomarkers in sediments from OD1507-03TC-41GC-03PC 250
Prior to the analysis of organic constituents the samples were freeze-dried (-45 °C; 0.2 mbar; 48 h) at Aarhus University using a Christ Alpha 1-4 LSC freeze drier. The dried samples were weighed and dry bulk densities (DBD) were determined from the samples respective volume and dry weight. Subsequently, the samples were homogenized with a dichloromethane (DCM) cleaned pestle and mortar.

Total organic carbon (TOC) quantification and analysis of carbon isotopic composition of TOC (δ 13 Corg) 255
For TOC and δ 13 Corg, 10 mg of homogenized sample material was weighed into an Ag capsule (Elemental Microanalysis).
Depending on available material, duplicates or triplicates were prepared to test for reproducibility. Additionally, a soil standard (10 mg; Elemental Microanalysis Soil Standard (Sandy) OAS 133506; 0.76 % w/w TOC) and blanks were added every ~15 samples. To remove the inorganic carbon, 35% HCl was added one drop at a time until no further reaction was observed (approximately 4 drops per sample, standard, and blank). The samples were then dried on a hotplate (50°C) overnight. The Ag 260 capsules were folded and placed inside a Sn capsule (Elemental Microanalysis), which was packed tightly and stored in a desiccator until analysis.
The δ 13 Corg and % wt TOC were measured using a continuous-flow IsoPrime IRMS coupled to an Elementar PyroCube elemental analyser at the Aarhus AMS Centre (AARAMS), Aarhus University, Denmark. δ 13 Corg is reported in ‰ versus Vienna Pee Dee Belemnite (VPDB). An in-house standard Gel-A was used as primary standard yielding ±0.2‰ and ±0.3‰ 265 for carbon and nitrogen analysis respectively. Further, secondary in-house and international standards were used to check the normalization to the VPDB. The mean reproducibility using 10 duplicate samples is ±0.03% and ±0.4‰ for % wt TOC and δ 13 Corg respectively. For both acid pre-treated and non-acid pre-treated Elemental Microanalysis Soil Standard (Sandy) OAS 133506 samples the reproducibility of 7 samples is ±0.02% wt TOC. The δ 13 Corg reproducibility of non-acid pre-treated Elemental Microanalysis Soil Standard (Sandy) OAS 133506 samples is ±0.1‰ whereas acid pre-treated samples show a mean 270 reproducibility of ±0.5‰. TOC concentrations are reported in weight % (wt.%) and TOC fluxes (µg cm -2 yr -1 ) are derived using the samples individual DBD and linear sedimentation rates (LSR), which were calculated using the datums from 14 C analysis of benthic and planktonic foraminifera (Reilly et al., 2019).
According to polarity, the different lipid classes were separated using silica column chromatography. Nonpolar lipids, such as 280 IP25 and HBI II were eluted with hexane, while the more polar sterols were eluted with DCM:MeOH (1:1, v/v) (~7 mL, each).
All biomarker samples were analysed at Aarhus University using an Agilent 7890B GC fitted with an HP-5ms Ultra Inert column (30 m x 250 μm x 0.25 μm) coupled to a 5977A series mass selective detector and equipped with a Gerstel multipurpose sampler (MPS). Prior to analysis HBI and sterol extracts were diluted with 50 µL hexane (using the MPS system) and 0.500 mL DCM (manually), respectively. For GC-MS operating conditions see Supplementary Table 2 Table 2) to the PA of the respective internal standard (Belt et al., 2012) under consideration of an instrumental response factor (based on the reference sediment) and the mass or the TOC concentrations of the sediment extracted (Belt et al., 2012). Due to unknown concentrations of dinosterol in the reference sediment, an individual response factor could not be determined for this compound. Instead, the 295 average response factor for brassicasterol and cholesterol was used. Thus, while the relative dinosterol concentrations and trends hold true, absolute concentrations might not be accurate. In addition to the biomarker concentrations in ng g -1 of dry sediment (ng g -1 sed) and µg g -1 TOC, biomarker fluxes were calculated using the samples individual DBD and linear sedimentation rates (LSR). Fluxes are reported in ng cm -2 yr -1 and are interpreted alongside biomarker concentrations to avoid bias related to jumps in the LSR. 300

Planktonic and benthic foraminiferal abundances
The OD1507-03TC-41GC-03PC benthic and planktonic foraminiferal abundances were determined on 58 samples. The benthic foraminiferal counts include both calcareous and agglutinated species. Where sufficient core material was available the sample depths correspond to the TOC and biomarker samples. The foraminiferal samples were weighed and wet sieved at 63 µm. The >63 µm fraction was counted wet, submerged in a 'storage' solution of 70 % distilled water and 30 % ethanol with 305 baking soda to preserve fragile calcareous and agglutinated tests. A wet splitter was used when necessary to achieve a count of at least 200-300 benthic foraminifera. Planktonic foraminifera were counted in the benthic split. Equivalent dry weights of the foraminiferal samples were calculated using the wet weights of the samples and the wet and dry weights of other samples from the same depths. That way, the numbers of benthic and planktonic foraminifera per gram of dry sediment could be calculated without drying the samples. Foraminiferal fluxes were calculated using the samples DBD and LSR and reported in 310 specimens cm -2 yr -1 , where the depth of the biomarker and foraminiferal samples were not identical, the DBD was linearly interpolated between neighbouring biomarker samples.

Results
The results of organic constituents and foraminiferal abundances in OD1507-03TC-41GC-03PC are presented on depth, to account for the bottom ca. 1.5 m of the core, where sediment ages are unconstrained (Reilly et al., 2019). This also allows to 315 compare the data to the sedimentary facies in OD1507-03TC-41GC-03PC, characteristic of the glacial dynamics in the fjord.   The ratio of HBI II and IP25 (DIP25) has previously been used as an indicator for sea surface temperature (SST) and thus as a tracer of warmer water masses (Hörner et al., 2016;Xiao et al., 2013), in line with higher temperatures being more favourable for the synthesis of double bonds. Other studies, however, did not find a relationship between DIP25 and SST and propose 350 instead that a steady DIP25 reflects stable sea-ice conditions, while a variable DIP25 indicates more unstable sea-ice conditions (Belt and Müller, 2013;Cabedo-Sanz et al., 2013). In outer Petermann Fjord, the DIP25 ratio is low throughout sedimentary unit 3 and rises before the unit3/unit 2 boundary ( Supplementary Fig. 4). This is followed by the highest recorded DIP25 values between 430 cm and 520 cm in unit 2. At 430 cm a sharp decline in DIP25 values marks the onset of a 110 cm long section with relatively low but variable DIP25 values (σ 2 = 0.41). At 320 cm, a small increase in DIP25 is followed by a relatively steady 355 decline throughout upper unit 1C, unit 1B, and unit 1A ( Supplementary Fig. 4). This coincides with reduced variance in the data (σ 2 = 0.26).

Sterol biomarker concentrations
Sterol biomarkers were determined on 94 samples with a depth and temporal resolution of 5.7±2.3 cm and 102±51 years, respectively. We measured brassicasterol, dinosterol, and cholesterol (hereinafter grouped as marine sterols) and campesterol 360 and β-sitosterol (hereinafter grouped as terrestrial sterols). All sterols were present consistently throughout the core (Table 3).
There are minor differences in the temporal evolution of sterol concentrations normalized to the amount of sediment extracted and to the TOC content of the samples ( Supplementary Fig. 3). These are especially apparent in the dinosterol, campesterol, and β-sitosterol concentrations. The differences are focused on the lower part of the record (400-500 cm), where Petermann Fjord is influenced by enhanced influx of terrestrial organic carbon (see section 5.1). Thus, we use sterol concentrations 365 normalized to the extracted sediment mass and the resulting fluxes to make inferences about past environmental changes. One sample at ~319 cm has β-sitosterol concentrations three times higher than the average, while all other sterol concentrations fall

385
Similar to the HBIs, the concentrations of all sterols increase at the transition from sedimentary units 3 to 2 with the most significant increase observed in β-sitosterol and campesterol (Fig. 4B, C). Both terrestrial sterols are characterized by a double peak throughout sedimentary unit 2, followed by a decline at the 1C/2 boundary, while the concentrations of marine sterols remain relatively stable throughout this interval (Fig. 4). Throughout unit 1C an increase in the cholesterol and dinosterol concentrations is observed from 350 cm, while their fluxes, alongside all other sterol concentrations remain low until 260 cm 390 ( Fig. 4D, E). Peak concentrations are reached between 199 cm and 189 cm, followed by a slight decrease in all sterol concentrations and a recovery during lower sedimentary unit 1B (Fig. 4). Simultaneously, sterol fluxes are at their highest between 94 cm and 194 cm, corresponding to the interval of maximum LSR (Fig. 4). While the sterol fluxes decrease from 94 cm, concentrations peak from 74-84 cm, followed by a sharp decrease prior to the transition from sedimentary units 1B to 1C.
Unit 1C is characterized by overall low sterol concentrations and fluxes (Fig. 4). 395 For further insight into the environmental factors driving sterol variability, the marine sterol index (sum of marine sterols/sum of all sterols) was determined (Stein et al., 2017). This indicates that sedimentary units 3, 2, and the lowermost unit 1C have a higher relative concentration of terrestrial sterols, with a slight increase in the relative concentration of marine sterols just prior to the 2/1C boundary (Fig. 4A). The relative concentration of marine sterols increases between 314 cm and 59 cm, followed by a decrease at the boundary of sedimentary units 1A/1B (Fig. 4A). 400 The overall amount of TOC in OD1507-03TC-41GC-03PC is very low, varying between 0.1 wt.% and 0.3 wt.% (Fig. 3C).
Between 289-394 cm and at 109 cm, samples recorded high TOC (0.3-1.8 wt.%) in combination with low δ 13 Corg (-26.2-30.4 405 ‰), suggesting incomplete inorganic carbonate removal during sample processing. The interval between 289 cm and 394 cm is associated with high Ti/Ca ratios and IRD delivery to the core site, indicating increased glacial sedimentation in outer Petermann Fjord (Reilly et al., 2019) (Fig. 3A), which might have been associated with input of detrital carbonates. Thus, these samples have been removed from the TOC and δ 13 Corg record ( Fig. 3B and C).
The lowest TOC is encountered in sedimentary unit 3, followed by a sharp increase prior to the boundary of sedimentary units 410 3 and 2 at 520 cm depth. The lower unit 2 is characterized by large variability in the TOC content with two peaks at 520 cm and 490 cm, while the upper sediments of the same unit have a more stable TOC content around 0.15 wt.% (Fig. 3C). The δ 13 Corg, measured on the same material as TOC, varies between -24.5 ‰ and -28.2 ‰. The lowest values are observed in sedimentary units 3 and 2 with minima at 490 cm and 530 cm and a relatively steady rise throughout the latter (Fig. 3B).

Benthic and planktonic foraminiferal concentrations
The benthic and planktonic foraminiferal abundances vary between 0 and 429 specimens g -1 sediment and 0 and 92 specimens g -1 sediment, respectively (Fig. 3D, E). This corresponds to benthic and planktonic fluxes of 3-32 specimens cm -2 year -1 and 0-8 specimens cm -2 year -1 (Fig. 3D, E). In sedimentary units 2 and 3, benthic and planktonic foraminiferal abundances are low. 425 Both benthic and planktonic abundances increase near the unit 2/unit 1C boundary, at 404 cm and 384 cm depth, respectively. Unit 1C is characterized by the highest overall foraminiferal abundances and fluxes with planktonic abundance and fluxes peaking between 304 cm and 312 cm prior to the benthic abundance, which peaks at 254 cm (Fig. D, E3). Planktonic foraminifera abundances and fluxes decline from 250 cm in unit 1C and continue at low values through unit 1B and are nearly absent in unit 1A (Fig. 3E). Benthic foraminiferal abundances and fluxes are more variable, but reach their lowest abundance 430 in unit 1A (Fig. 3 D).

Holocene variability in organic carbon sources and sea-ice dynamics in Petermann Fjord
Organic carbon (Corg) in Petermann Fjord is derived from a variety of sources, including in situ pelagic and sympagic production, advection of marine organic matter, and input of both modern and ancient Corg from the surrounding landmasses. 435 While HBI concentrations are only marginally influenced by the addition of terrestrial Corg, sterols are important constituents in both marine and terrestrial primary producers (Belicka et al., 2004;Rontani et al., 2014;Volkman, 1986Volkman, , 2003Volkman et al., 1993Volkman et al., , 2000Yunker et al., 1995) (Table 2). Thus, in combination with δ 13 Corg, they can provide information on the dominant Corg source.
Sedimentary unit 3 represents grounding zone proximal sedimentation, associated with the deglacial retreat of PG into the 440 fjord (Jakobsson et al., 2018;Reilly et al., 2019). This is in line with very low biomarker and TOC concentrations, and near absence of benthic and planktonic foraminifera (Fig. 5A-D), suggesting reduced primary productivity under an ice shelf/thick sea-ice cover and/or high sedimentation rates close to the grounding zone diluting the concentrations of organic constituents and marine microfossils close to the margin of a retreating glacier. Both sedimentary unit 3 and 2 are beyond the lowermost available radiocarbon age, which is why the influence of sedimentation rates on the biomarker and TOC concentrations cannot 445 be assessed. The low marine sterol index throughout sedimentary unit 3 indicates that the Corg input was likely dominated by terrestrial sources, supported by low δ 13 Corg values, characteristic of terrestrial vegetation in high latitudes (-26 ‰ to -28 ‰) (Ruttenberg and Goñi, 1997) (Fig. 3B). Where Washington and Hall Land border Petermann Fjord they are characterized by sedimentary rocks of lower Paleozoic age, primarily composed of shallow marine carbonates and evaporites (Dawes et al., 2000;Harrison et al., 2011). These contain ancient biomass that can be traced in marine sediments. The thermal maturity of 450 these rocks, however, indicates that fossil sterols will have been, to a large part, degraded to steranes (Parnell et al., 2007), not measured as part of this study. Thus, the increased abundance of terrestrial sterols in sedimentary unit 3 is most likely related to input of fresh terrestrial organic material. Campesterol and β-sitosterol are dominant in vascular plants (such as herbs), but are also found in mosses and lichens (Matsuo and Sato, 1991;Safe et al., 1975), common to the high Arctic tundra. The deglacial retreat of PG occurs relatively late during the deglaciation of northern Greenland (<7,600 cal yrs BP) (Jakobsson et 455 al., 2018), falling into the regional Holocene Thermal Maximum (HTM; ca. 11,000-5,500 cal yrs BP) (Kaufman et al., 2004) when adjacent Washington Land was already deglaciated (Ceperley et al., 2020) and pollen records from northern Greenland and Ellesmere Island suggest a more productive terrestrial Arctic ecosystem (Gajewski, 2015;Mode, 1996). This supports enhanced input of fresh terrestrial Corg to Petermann Fjord in response to glacial erosion and enhanced meltwater drainage of surrounding landmasses during the deglacial retreat of PG. 460  Mapped submarine landforms suggest that the retreat of PG into the fjord was rapid, driven by ice cliff instability and promoted 475 by the retrograde slope of the outer fjord sill (Jakobsson et al., 2018). An inner sill (Tinto et al., 2015) ~30 km seaward of the present day grounding zone (Reilly et al., 2019) likely halted the retreat, allowing for the formation of an extensive ice tongue associated with the laminated, IRD-poor lithofacies of unit 2 (Reilly et al., 2019). The transition from units 3 to 2 is marked by an increase in all biomarkers, TOC, and benthic and planktonic foraminiferal abundances (Fig. 5), suggesting an overall increase in Corg input to outer Petermann Fjord and/or lower sedimentation rates in response to the larger distance of the core 480 site from the grounding zone. The marine sterol index remains low, suggesting continued dominance of terrestrial versus marine sterols, while increasing δ 13 Corg values indicate enhanced importance of marine Corg sources to Petermann Fjord. Thus, compared to sedimentary unit 3, the organic matter in sedimentary unit 2 was likely derived from more varied sources, including in situ/advected marine Corg and terrestrial Corg. This supports a marginal/sub-ice tongue regime in outer Petermann Fjord with no or only intermittent sympagic and pelagic primary productivity in the fjord and low food supply to sustain 485 benthic productivity (Jennings et al., 2020).

line with dots) normalised to the amount of extracted sediment (ng g -1 sed.). (d) IP25 fluxes (blue, filled in area) and absolute concentration (blue line with dots) normalised to the amount of extracted sediment (ng g -1 sed.). (e) Marine sterol index (sum of marine sterols/ sum of all sterols) (purple line). (f) TOC fluxes (green, filled in area) and TOC concentrations (wt. %, green line). (g) XRF Ti/Ca ratios (grey) and the CT >2 mm clast index (brown) (Reilly et al., 2019). (h) Reconstruction of the Holocene ice tongue extent (minimum extent in dark blue, maximum extent in light blue) in Petermann
The early Holocene ice tongue of PG collapsed around 6,900 cal yrs BP (unit 2/unit 1C boundary), marked by the abrupt appearance of IRD clasts in sediments across Petermann Fjord (Reilly et al., 2019) (Fig. 5G). This is associated with an increase in the marine sterol index and TOC (Fig. 5), suggesting an increase in the Corg delivery to outer Petermann Fjord and relatively more input of marine versus terrestrial sterols compared to previous sedimentary units. A second, larger, increase in the marine 490 sterol index is evident at 5,100 cal yrs BP (Fig. 5E), contemporaneously with decreasing XRF Ti/Ca ratios and IRD flux (Reilly et al., 2019). This indicates reduced glacial erosion and calving activity or a decrease in the delivery of erosional products to outer Petermann Fjord from 5,100 cal yrs BP and dominant input of marine over terrestrial Corg, supported by δ 13 Corg (Fig. 5,   Fig. 3). The marine sterol index remains high until ~600 cal yrs BP, while δ 13 Corg is high throughout the rest of the core, suggesting primarily delivery of marine-derived organic matter to outer Petermann Fjord during the mid-to-late Holocene. 495 The break-up of the early Holocene ice tongue is also associated with a small increase in the sedimentary IP25 and HBI II concentrations, as well as an increase in the benthic and planktonic foraminiferal abundances. Between 6,900 cal yrs BP and 5,500 cal yrs BP, IP25 fluxes are low but variable, indicating an unstable/variable sea-ice regime with predominantly reduced seasonal sea-ice concentration during spring and low rates of sympagic productivity. An unstable sea-ice regime is further supported by variable DIP25 values (Supplementary Fig. 4). Alternatively, increased meltwater runoff related to the collapse 500 of the ice tongue and associated retreat of PG, might have caused low/variable IP25 fluxes by creating environmental conditions unfavourable for IP25-producing diatom species during times of increased freshwater discharge (Ribeiro et al., 2017). The simultaneous increase in TOC, benthic, and especially planktonic foraminiferal fluxes, however, supports a reduced seasonal sea-ice cover and a shift from a regime marginal to or below an ice tongue towards ameliorated conditions with a prolonged open water season and enhanced pelagic primary productivity in the fjord, allowing planktonic foraminifera to thrive (Fig.  505   5B). A reduced seasonal sea-ice cover is further supported by evidence of low sea-ice concentrations in Hall Basin until at least 6,000 cal yrs BP (Jennings et al., 2011b). Regional records from northern Greenland and the western Canadian Arctic Archipelago (CAA) demonstrate a HTM between 11,000 cal yrs BP and 5,500 cal yrs BP (Belt et al., 2010;Briner et al., 2016;England et al., 2008;Funder et al., 2011;Jennings et al., 2011b;Kaufman et al., 2004;Knudsen et al., 2008;Lecavalier et al., 2017;Ledu et al., 2010;Vare et al., 2009), associated with regional mean annual surface air temperatures 3±1°C higher than 510 pre-industrial (1750 CE) (Lecavalier et al., 2017). Thus, reduced seasonal sea-ice concentrations in Petermann Fjord between 6,900 cal yrs BP and 5,500 cal yrs BP likely represent the late stages of the HTM in the northern Nares Strait region.
From ca. 5,800 cal yrs BP the increase in planktonic and benthic foraminiferal fluxes steepens, associated with less variable but still low IP25 fluxes from 5,500 cal yrs BP (Fig. 5A, B, and D). While benthic foraminifera respond to sea-ice changes via its influence on marine productivity and the resulting changes in food supply to the seafloor (Seidenkrantz, 2013), planktonic 515 foraminiferal abundances have been shown to be highest in the open water region and along the ice margin, with only few individuals occurring under persistent sea ice (Carstens et al., 1997;Mayot et al., 2020;Pados and Spielhagen, 2014). In the modern environment of the northern Nares Strait and Petermann Fjord, planktonic foraminiferal abundances are very low in the fjord and much higher in the mobile sea ice regime of Nares Strait (Jennings et al., 2020), characterized by a shorter seasonal sea-ice season than outer Petermann Fjord (Fig. 2). Thus, high planktonic foraminiferal fluxes between 5,800 cal yrs 520 BP and 3,600 cal yrs BP suggest sustained periods of seasonally open waters during summer, while the continuously low IP25 fluxes are consistent with reduced seasonal sea-ice concentrations and the absence of a sympagic spring bloom in outer Petermann Fjord (Fig. 5 D). Alternatively, IP25 production might have been suppressed due to lasting meltwater discharge into the fjord, resulting from the influence of the retreating PG on the outer fjord environment. However, both the >2 mm clast index and the XRF Ti/Ca ratios decrease throughout this interval (Reilly et al., 2019), suggesting a gradual reduction of the 525 influence of glacial activity on the outer fjord (Fig. 5G).
From 3,900 cal yrs BP, IP25 fluxes increase steeply accompanied by an increase in all sterol fluxes (Fig. 4, Fig. 5D). This suggests a shift in the ecosystem in outer Petermann Fjord, associated with a transition from a regime dominated by pelagic primary productivity towards a regime characterized by enhanced sympagic productivity. Especially after 3,600 cal yrs BP, rapidly decreasing planktonic foraminiferal fluxes and steadily increasing IP25 fluxes indicate a progressively longer sea-ice 530 season and enhanced sympagic productivity in outer Petermann Fjord (Fig. 5B, D). This falls into a period of long-term declining regional atmospheric temperatures recorded at Agassiz ice cap (Lecavalier et al., 2017) and in lake records in NW Greenland (Axford et al., 2019;Lasher et al., 2017). Neoglacial cooling has been observed in numerous marine and terrestrial archives around Greenland and the wider North Atlantic region (e.g. England et al., 2008;Hansen et al., 2020;Jennings et al., 2011a;Limoges et al., 2020;Vare et al., 2009), as a response to decreasing northern hemisphere summer insolation (Marcott 535 et al., 2013). Thus, enhanced seasonal sea-ice conditions in Petermann Fjord from 3,900 cal yrs BP (Fig. 5D), likely record the onset of Neoglacial cooling in the northern Nares Strait region.
Peak IP25 fluxes around 2,500 cal yrs BP are associated with high fluxes of marine and terrestrial sterols as well as benthic foraminifera, while planktonic foraminiferal fluxes are low (Fig. 4, Fig. 5A, B, D), indicating a prolonged seasonal sea-ice cover with only short periods of open water during summer. From 2,500-2,100 cal yrs BP a two-stepped decrease in IP25 fluxes 540 is observed, of which the second decline is accompanied by a decrease in the TOC, benthic foraminiferal flux, and a small decrease in the (already low) planktonic foraminiferal flux (Fig. 5A, B, F). A contemporaneous decline in all productivity indicators and sea-ice biomarkers, suggests a restriction in the pelagic and sympagic primary productivity alike, most likely as a response to further lengthening of the sea-ice season to near-perennial sea-ice cover. This interval precedes the late Holocene inception of a small ice tongue in Petermann Fjord at 2,200-2,100 cal yrs BP, inferred from Ti/Ca ratios and the stacked >2 545 mm clast index from four cores in Petermann Fjord (Reilly et al., 2019). After 2,100 cal yrs BP, sea-ice biomarkers, TOC, and benthic foraminiferal fluxes recover, while the planktonic foraminiferal abundance remains low (Fig. 5A, D, F), indicating an ecosystem dominated by sympagic productivity and enhanced seasonal sea-ice cover, similar to the interval 3,600-2,500 cal yrs BP. Compared to the deglacial PG ice tongue, the late Holocene ice tongue (<2,100 cal yrs BP) (Reilly et al., 2019) does not seem to be associated with increased input of terrestrial organic matter to outer Petermann Fjord (Fig. 5). A possible 550 explanation could be lower atmospheric temperatures compared to the early Holocene (Lecavalier et al., 2017), associated with a less diverse and more sparse terrestrial flora in the high Arctic (Gajewski, 2015) and decreased meltwater input. Another reason could be increased distance of OD1507-03TC-41GC-03PC from the PG grounding zone during the late Holocene (Reilly et al., 2019), resulting in reduced delivery of meltwater-derived Corg.
Around 1,300 cal yrs BP a sharp decline in IP25, sterol, and TOC fluxes is observed, while the concentration of IP25 decreases 555 more gradually and the concentrations of sterols and TOC increase (Fig. 5). The sharp decrease in fluxes at this time coincides with a large drop in the LSR (Fig. 3A), which might be biasing the flux data. Instead, the decline in biomarker fluxes at 950 cal yrs BP, accompanied by a decrease in the biomarker concentrations, seems to be a more reliable feature (Fig. 5). This is associated with declining benthic foraminiferal fluxes, and followed by decreases in the planktonic foraminiferal and TOC fluxes at 700 cal yrs BP (Fig. 5A, B, F), suggesting a return to near-perennial sea-ice conditions with reduced pelagic and 560 sympagic primary productivity, similar to the interval from 2,500-2,100 cal yrs BP. At ca. 600 cal yrs BP a rapid extension of the Petermann ice tongue to its modern limits (Reilly et al., 2019), resulted in (at least intermittent) cover of the core site, in line with low biomarker and foraminiferal fluxes between 600 cal yrs BP and the top of the core (Fig. 5). The latter indicates low rates of primary productivity in a fjord, which is nearly completely covered by an ice tongue. This is further supported by a decrease in the marine sterol index between 600 cal yrs BP and the top of the core, suggesting a relative decrease in the input 565 of marine organic matter (Fig. 5E). However, while there is a small decrease in the TOC at this time, there is no corresponding decrease in the δ 13 Corg (Fig. 3B), suggesting that marine Corg was still the main source of TOC in outer Petermann Fjord.

Nares Strait sea-ice dynamics over the last 7,000 cal yrs BP
Thus far, only one other biomarker-based sea-ice reconstruction, from station Kane2b in north-western Kane Basin, exists in Nares Strait for comparison with our records (Fig. 1) (Georgiadis et al., 2020). We propose that depending on the ice arch 580 configuration in Nares Strait, Kane Basin and outer Petermann Fjord likely experience opposing conditions related to the proximity of the ice edge during spring/early summer ( Supplementary Fig. 1), which is the dominant productivity season of sea-ice biomarkers. Kane2b is located near/under the southern ice arch in Nares Strait. Thus high sympagic productivity and IP25 fluxes occur during times of a stable ice arch in Smith Sound and sea-ice edge conditions exist during spring/summer (Georgiadis et al., 2020) (Fig. 6B, D). Georgiadis et al. (2020) interpret HBI III as generally indicating ice-laden/fresh surface 585 waters, which can occur following the break-up of the southern ice arch and during times of mobile sea ice in Nares Strait.
Thus, the southern ice arch scenario is likely associated with variable HBI III fluxes, depending on the seasonal timing of ice arch break-up. During the southern ice arch scenario, sea ice in Petermann Fjord does not break up until late summer/early autumn (Fig. 2), likely hindering a pronounced in-ice bloom related to MIZ conditions during spring/summer, resulting in relatively low sea-ice and phytoplankton biomarker concentrations (Fig. 6B, D). In years when only the northern ice arch 590 forms, sea ice formed locally in Nares Strait either remains mobile throughout the winter or breaks up during early spring.
Under these conditions HBI III fluxes at Kane2b can be relatively high depending on the sea ice flux through Nares Strait during spring/summer, but no pronounced sympagic spring bloom occurs in Kane Basin (Fig. 6C). Instead, outer Petermann Fjord experiences spring MIZ conditions, as the formation of landfast ice in Petermann Fjord is independent of the formation of landfast ice in Nares Strait (Fig. 2, Supplementary Fig. 1). This is likely associated with a significant spring sympagic bloom 595 in outer Petermann Fjord and enhanced primary productivity related to the vicinity of the ice edge, resulting in increased concentrations of sea-ice and primary productivity biomarkers in outer Petermann Fjord (Fig. 6C). Periods with contemporaneously low concentrations of sea-ice and primary productivity biomarkers in Kane Basin and Petermann Fjord, point towards low spring sea-ice concentration in the entire Nares Strait, likely associated with a failure of both ice arches (Fig. 6A). 600 In addition to the Kane Basin record, sea-ice dynamics around Ellesmere Island and NE Greenland have been inferred from records of driftwood delivery and beach ridge formation (England et al., 2008;Funder et al., 2011), providing context with respect to Holocene sea-ice conditions in the Arctic Ocean and Lincoln Sea. Driftwood is transported with Arctic multiyear ice and is deposited along the coastlines of northern Greenland and Ellesmere Island when landfast ice breaks up during summer (Funder et al., 2011). The formation of beach ridges also depends on sufficient wave action and open water along the 605 coast. Thus, abundant driftwood delivery to Ellesmere Island/north-eastern Greenland together with abundant formation of beach ridges is indicative of seasonally open waters along the coast (Fig. 6A). Sea-ice conditions with year-round landfast ice along the coast, on the other hand, will result in little or no driftwood landings and reduced formation of beach ridges (Fig.   6D). Lastly, variable ice conditions in Lincoln Sea result in occasional/little driftwood landings and reduced/variable formation of beach ridges (Fig. 6B, C). 610 While the sediment core from station Kane2b covers the last 9,000 cal yrs BP (Georgiadis et al., 2020) (Fig. 1), Petermann Fjord did not deglaciate until ~7,600 cal yrs BP (Jakobsson et al., 2018) and the age model for OD1507-03TC-41GC-03PC is only constrained for the last 7,000 cal yrs BP. Thus, we focus on the interval from 6,900 cal yrs BP to the present. Between 6,900 cal yrs BP and 5,500 cal yrs BP, corresponding to the late stages of the regional HTM (Kaufman et al., 2004), sea-ice biomarker and foraminiferal fluxes in Petermann Fjord and Kane Basin (Georgiadis et al., 2020), indicate reduced seasonal 615 sea-ice occurrence (Fig. 7F, G). This corresponds to sparse driftwood delivery but abundant beach ridge formation around NE Greenland from 8,500-6,000 cal yrs BP, suggesting a minimum in Arctic multi-year ice and seasonally open water along the coast (Funder et al., 2011;Möller et al., 2010). Seasonally open water is further supported by maximum driftwood delivery to Disraeli Fiord and Clements Markham Inlet (CMI) on Ellesmere Island (England et al., 2008) (Fig. 1, Fig. 7E). This is in line with reduced seasonal sea-ice occurrence in Nares Strait, suggesting predominantly mobile sea ice with no or only occasional 620 ice arch formation and export of Arctic sea ice through Nares Strait (Fig. 6A) (Georgiadis et al., 2020).

640
At 5,500 cal yrs BP increasing sea-ice biomarker fluxes in Kane Basin indicate a shift towards later sea-ice retreat and ice edge productivity, interpreted to reflect recurrent formation of the southern ice arch in Smith Sound/Kane Basin between 5,500 cal yrs BP and 3,000 cal yrs BP (Georgiadis et al., 2020) (Fig. 7F). A study of seabird colonies in the NOW region, suggests the arrival of little auk colonies in NW Greenland at 4,400 cal yrs BP, supporting the opening of the NOW lee of the Smith Sound ice arch during spring/summer (Davidson et al., 2018) (Fig. 7B). Little auk are zooplanktivore, feeding on the abundant 645 copepod population of the NOW. Thus, large colonies of little auk in Greenland are only found in vicinity of productive polynyas, where open water is available for foraging upon their arrival in spring (Davidson et al., 2018). Productive and strong NOW conditions from 4,400 cal yrs BP are also inferred based on foraminiferal assemblages from the central polynya region (Jackson et al., 2021). Simultaneously, landfast sea ice started to form in Disraeli Fiord on northern Ellesmere Island from ~5,500 cal yrs BP, covering large parts of the coast by ~3,500 cal yrs BP (England et al., 2008), and shorter periods of open 650 water and restricted beach ridge formation occur around NE Greenland (Funder et al., 2011). Even though this suggests an intensification of landfast sea-ice in the Lincoln Sea and the Arctic Ocean, open water along the NE Greenland coast was more abundant compared to today until at least 4,500 cal yrs BP (Funder et al., 2011). The period from 5,500-3,500 cal yrs BP, thus marks the transition from early Holocene warmth to Neoglacial cooling in the wider Nares Strait region. Compared to Kane2b, biomarker fluxes do not increase significantly at 5,500 cal yrs BP in outer Petermann Fjord (Fig. 7G). Instead, benthic and 655 planktonic foraminiferal fluxes are at their highest between 5,500 cal yrs BP and 3,500 cal yrs BP (Fig. 6A, B), interpreted to reflect reduced seasonal sea-ice concentrations, the absence of a sympagic spring bloom, prolonged periods of seasonally open waters during summer, and increased pelagic primary productivity. Alternatively, the continuous influence of the retreating PG during this interval, potentially associated with increased meltwater discharge into the fjord, might have caused low concentrations of sea-ice biomarkers in outer Petermann Fjord (Ribeiro et al., 2017). Nonetheless, the high foraminiferal fluxes 660 indicate conditions that differ from what would be expected in outer Petermann Fjord under a stable southern ice arch scenario (Fig. 6B). A possible reason for this could be enhanced seasonality, in particular enhanced winter cooling, due to the increasing sea-ice extent in the Arctic Ocean (England et al., 2008;Funder et al., 2011), while summer insolation was still relatively high (though decreasing) compared to the late Holocene (Fig. 7). Modelling studies have shown that the loss of sea ice in the Arctic Ocean during the early Holocene counteracted the increased seasonality prescribed by insolation forcing, due to enhanced 665 ocean-atmosphere heat flux during winter (Fischer and Jungclaus, 2011). The Arctic Ocean thus acted as a heat reservoir with increased latent and sensible heat flux during winter as a result of the reduced sea-ice cover (Fischer and Jungclaus, 2011).
Conversely, decreasing summer insolation and the associated increase in Arctic Ocean sea-ice cover during the mid-Holocene ( Fig. 7) strengthened the insulating effect of sea ice on the ocean and led to pronounced cooling during autumn/winter (Fischer and Jungclaus, 2011). Enhanced seasonality could explain the observed proxy patterns in Nares Strait between 5,500 cal yrs 670 BP and 3,500 cal yrs BP, with an early seasonal break-up of the southern ice arch, leading to increased pelagic primary productivity in northern Nares Strait. This way, the spring ice edge bloom might still have occurred in the southern Nares Strait, followed by a rapid sea-ice retreat and open water conditions during summer/autumn, as recorded in outer Petermann Fjord. An early seasonal break-up of the southern ice arch is also supported by high HBI III fluxes, especially between 4,500 cal yrs BP and 3,500 cal yrs BP (Georgiadis et al., 2020). Thus, we suggest that this interval represents the transition from 675 reduced sea-ice conditions and the lack of ice arches in Nares Strait during the late HTM towards more stable sea-ice conditions associated with the seasonal formation of a recurrent southern ice arch from at least ~4,400 cal yrs BP.
From 3,900 cal yrs BP, increasing IP25 fluxes in Petermann Fjord, attest to a shift in the primary productivity regime from predominantly pelagic to MIZ/sympagic primary productivity at 3,500 cal yrs BP. A similar increase in IP25 concentrations from ~4,000 cal yrs BP is also observed in the western CAA, in Barrow, Victoria, and Dease Strait (Belt et al., 2010;Vare et 680 al., 2009), in line with enhanced seasonal sea-ice formation and the onset of regional Neoglacial cooling. Conversely, IP25 concentrations in Kane Basin decrease from ~4,000 cal yrs BP until 1,400 cal yrs BP (Fig. 7F), suggesting no stable southern ice arch in Smith Sound. Instead, the driftwood delivery to CMI ceases at 3,500 cal yrs BP (England et al., 2008). This likely indicates a regime shift in Nares Strait with the northern ice arch becoming more prominent between 3,500 cal yrs BP and 1,400 cal yrs BP (Fig. 6C). In years where only the northern ice arch forms, along strait winds in Nares Strait keep locally 685 formed sea ice mobile year-round (Vincent, 2019), creating MIZ conditions in Petermann Fjord during spring/summer ( Supplementary Fig. 1), in line with high IP25 concentrations. Between 2,500 cal yrs BP and 2,100 cal yrs BP biomarker fluxes in outer Petermann Fjord, suggest a shift towards near-perennial sea-ice conditions (Fig. 7G). This is associated with some driftwood landings in CMI (England et al., 2008) and a period of no driftwood or beach ridge occurrences in NE Greenland, suggesting perennial landfast ice (Funder et al., 2011). The numbers of little auks at Qeqertaq (Salve Ø) are variable during 690 this time period, with a local peak in abundance at ~2,200 cal yrs BP (Davidson et al., 2018). In Kane Basin, IP25 fluxes are relatively low until ca. 1,400 cal yrs BP, however a minor peak (more pronounced in IP25 concentrations) can be observed between ca. 2,300 cal yrs BP and 2,000 cal yrs BP. This coincides with a period of low HBI III fluxes from ca. 2,700-2,100 cal yrs BP at Kane2b (Georgiadis et al., 2020). Georgiadis et al. (2020) interpret HBI III as indicating ice loaded fresh surface waters in relation to mobile sea ice/ice arch break-up in Nares Strait. A decrease in HBI III fluxes and concomitant small 695 increase in IP25 could thus result from intermittent formation of the southern ice arch in Kane Basin and opening of the NOW, consistent with the local abundance peak of little auks and sea-ice records from outer Petermann Fjord and north-eastern Greenland. Foraminiferal assemblages from the central NOW region (91-039-008P, Fig. 1) demonstrate a shift towards increased bottom water ventilation around 2,600 cal yrs BP (Knudsen et al., 2008), which could point towards strengthened polynya conditions and intensified brine formation in the NOW region. Unfortunately, however, no sediments were recovered 700 at site 91-039-008P between 2,450 cal yrs BP and 1,100 cal yrs BP (Knudsen et al., 2008). Increased bottom water ventilation is also inferred from measurements of total sulphur at CASQ1 (Fig. 1) (Jackson et al., 2021). Simultaneously, calcareous foraminiferal abundances are low between 2,500-2,000 cal yrs BP at CASQ1 (Fig. 1) (Jackson et al., 2021). While this hinders unambiguous environmental interpretations, it could also be a sign of strong polynya conditions and formation of CO2-rich brines, causing dissolution of the calcareous foraminiferal test and preservation of agglutinated species only (Jackson et al., 705 2021). Georgiadis et al. (2020) suggest that the period of mobile sea ice in Nares Strait between 3,500 cal yrs BP and 1,400 cal yrs BP was driven by a change in the dominant phase of the Arctic Oscillation (AO). The shift towards a positive AO between 3,000 cal yrs BP and 1,200 cal yrs BP (Darby et al., 2012) might have been associated with stronger along strait winds in Nares Strait. As ice arch formation in Nares Strait is a function of ice thickness, local wind stress, and atmospheric temperatures (Barber et al., 2001;Samelson et al., 2006), distinct atmospheric cooling spikes between ~2,500 cal yrs BP and 710 ~1,900 cal yrs BP, recorded at Agassiz ice cap (Lecavalier et al., 2017), might have been responsible for the enhanced sea-ice concentration and the formation of an intermittent southern ice arch in Nares Strait between 2,500 cal yrs BP and 2,100 cal yrs BP (Fig. 7A).
After 2,100 cal yrs BP and until 1,400 cal yrs BP, the combined biomarker records in Nares Strait support a return to a more northern ice arch dominated sea-ice regime. This is supported by foraminiferal assemblages from the central NOW region, 715 suggesting weaker polynya conditions from 1,800-1,400 cal yrs BP (Jackson et al., 2021). From 1,400 cal yrs BP onwards, IP25 fluxes in Kane Basin increase (Georgiadis et al., 2020), while they decline in Petermann Fjord (Fig. 7F, G) concomitant with high numbers of little auks at Qeqertaq (Salve Ø) (Davidson et al., 2018), suggesting a transition towards a more stable southern ice arch. These conditions are further consolidated from 950 cal yrs BP, where the sudden drop in sea-ice and primary productivity indicators in Petermann Fjord (Fig. 5, Fig. 7G) indicate near-perennial landfast sea ice in Nares Strait promoted 720 by the recurrent formation of a southern ice arch. Driftwood is absent in CMI from 2,200-400 cal yrs BP (England et al., 2008), suggesting that the northern arch might have been a stable feature as well (Fig. 7E, Fig. 6D). Periods of double ice arching in Nares Strait are particularly stable (Ryan and Münchow, 2017), in line with the coldest Holocene temperatures recorded at Agassiz ice cap throughout this interval (Lecavalier et al., 2017). At ~600 cal yrs BP a large jump in the extent of the late Holocene ice tongue in Petermann Fjord to its modern (pre 2010) limits, suggests a regime marginal to or under an ice tongue 725 in outer Petermann Fjord (Reilly et al., 2019) (Fig. 7H). Thus, further interpretations of the sea-ice environment in Nares Strait are hindered due to the influence of the ice tongue on primary productivity in Petermann Fjord.

Implications of sea ice for the formation and stability of the Petermann ice tongue
A recent study by Reilly et al. (2019) demonstrates that an extensive ice tongue in Petermann Fjord is a relatively recent feature compared to large parts of the Holocene. The final stage of the deglaciation in Petermann Fjord was characterized by the break-730 up of a deglacial ice tongue around 6,900 cal yrs BP, followed by a period of enhanced IRD and Ti/Ca ratios in the fjord sediments (6,900-3,500 cal yrs BP; Fig. 5G), indicating increased glacial sedimentation in outer Petermann Fjord. First at ~2,200 cal yrs BP a small ice tongue re-emerged, followed by gradual growth and a rapid expansion to its modern extent around 600 cal yrs BP (Reilly et al., 2019). Thus, the mid-to-late Holocene was marked by a ~4,700 year interval where no stable ice tongue existed in Petermann Fjord (Reilly et al., 2019). The presence of an ice tongue in Petermann Fjord is 735 determined by the interplay of several factors, including the surface mass balance of PG, the loss of glacial ice by calving, and the basal melt rates. Local changes in sea-ice dynamics have the potential to influence both the basal melt rates and the stability of the calving front at PG. The latter depends on the formation and seasonal duration of landfast sea ice in Petermann Fjord, creating an ice mélange that stabilizes the calving front and reduces the length of the calving season (Amundson et al., 2010;Carr et al., 2015;Robel, 2017;Todd and Christoffersen, 2014). Along with the increase in IP25 fluxes in outer Petermann Fjord, 740 indicating a transition towards enhanced sea-ice conditions, from at least 3,500 cal yrs BP, a cessation of the IRD flux is observed in OD1507-03TC-41GC-03PC (Fig. 5D, G) (Reilly et al., 2019). This suggests a gradual decrease in the calving of icebergs from PG during the mid-Holocene. We propose that this was partly in response to the formation of seasonal landfast ice in the fjord, shortening the calving season ( Fig. 6; Fig. 7). Sea ice also has the ability to influence the oceanic heat transport to Petermann Fjord, via modification to the circulation driven AAW inflow by regulating the wind stress at the 745 atmosphere/ocean interface in Nares Strait (Shroyer et al., 2017). Under modern conditions, AAW inflow is enhanced at times of mobile sea ice in Nares Strait, while landfast sea ice results in decreased oceanic heat flux to Petermann Fjord (Shroyer et al., 2017). In addition to Nares Strait sea-ice conditions, subglacial runoff and glacial isostatic uplift of the outer fjord sill likely also influenced the inflow of AAW to Petermann Fjord across the Holocene (Bennike, 2002;Cai et al., 2017;Washam et al., 2019). Recurrent formation of a southern ice arch and seasonal formation of landfast sea ice, may have persisted in Nares Strait 750 from 5,500-3,500 cal yrs BP, 2,500-2,100 cal yrs BP, and <1,400 cal yrs BP, based on the assessment of regional sea-ice records (England et al., 2008;Funder et al., 2011;Georgiadis et al., 2020, this study) (Fig. 6B, D). The interval from 5,500-3,500 cal yrs BP marks the transition from the HTM to Neoglacial cooling in the Nares Strait region. While records from southern Nares Strait suggest an opening of the NOW and spring ice edge conditions in Kane Basin from at least 4,400 cal yrs BP, records from outer Petermann Fjord indicate low sympagic but high pelagic productivity, suggesting prolonged periods of 755 open water during summer (Fig. 5A, B, D). Increased seasonality, with early break-up of the southern ice arch and late seasonal sea-ice formation might be able to explain the observed patterns (section 5.3). Thus, unlike today, the formation of a southern ice arch between 5,500 cal yrs BP and 3,500 cal yrs BP did likely not lead to a prolonged landfast ice season in Nares Strait.
Consequently, the longer mobile sea-ice season in Nares Strait might have contributed to AAW inflow to Petermann Fjord preventing the formation of an ice tongue throughout this interval. 760 Between 2,500 cal yrs BP and 2,100 cal yrs BP regional records of sea ice (England et al., 2008;Georgiadis et al., 2020, this study), seabird colonies (Davidson et al., 2018), and NOW dynamics (Jackson et al., 2021) suggest intermittent formation of a southern ice arch in Nares Strait, likely in response to regional atmospheric cooling spikes (Lecavalier et al., 2017). This interval spans the inception of a small late Holocene ice tongue in Petermann Fjord (Fig. 7H), suggesting that increased formation of landfast ice in Nares Strait may have reduced the oceanic heat flux to Petermann Fjord and stabilized the glacier 765 front. Together with atmospheric cooling this likely aided the formation of a small ice tongue at ~2,200 cal yrs BP (Reilly et al., 2019). Similarly, the interval <1,400 cal yrs BP including the rapid growth of the Petermann ice tongue to its modern limits at ca. 600 cal yrs BP (Reilly et al., 2019). Regional sea-ice records indicate the onset of a more recurrent southern arch from ~1,400 cal yrs BP, with a further decrease in sea-ice and primary productivity markers in Petermann Fjord and a concurrent increase in IP25 in Kane Basin at 950 cal yrs BP implying the formation of more stable ice arch conditions and potentially even 770 double ice-arching. Thus, both the inception of the ice tongue in Petermann Fjord at ~2,200 cal yrs BP and its rapid extension at 600 cal yrs BP are preceded by a transition towards stable southern ice arch conditions in the Nares Strait by 300-350 years ( Fig. 7, Fig. 6B, D). More severe sea ice, with longer periods of landfast ice in Nares Strait, thus likely promoted ice tongue growth in Petermann Fjord, either directly via stabilizing the calving front or indirectly via changes to the inflow in modified AW. In light of the recent development, our data suggest that the emerging dominance of the northern ice arch associated with 775 year-round mobile sea ice in Nares Strait and a shorter landfast sea-ice season in Petermann Fjord, will likely contribute to the destabilization of the ice tongue. Thus, a future reduction in landfast sea ice in Nares Strait and adjacent fjords would likely contribute to enhanced mass loss from the GrIS.

Conclusions
1. During the deglaciation (>6,900 cal yrs BP) outer Petermann Fjord experienced enhanced input of fresh terrestrial 780 Corg, likely due to the late deglacial retreat of PG and increased glacial erosion during the early HTM, characterized by more productive Arctic tundra vegetation in northern Greenland and Ellesmere Island.
2. Following deglacial retreat of PG and break-up of the deglacial ice tongue in Petermann Fjord at 6,900 cal yrs BP, sea-ice biomarkers and productivity indicators in outer Petermann Fjord suggest reduced seasonal sea-ice occurrence and predominantly pelagic primary productivity until 5,500 cal yrs BP, corresponding to the late stages of the regional 785 HTM. Together with other regional records, this suggests the lack of ice arches and export of sea ice from Lincoln Sea through Nares Strait.
3. The interval from 5,500-3,500 cal yrs BP marks the transition from the late stages of the HTM to the onset of Neoglacial cooling. This was likely associated with a formation of a southern ice arch and the opening of the NOW from at least ~4,400 cal yrs BP. However, an earlier seasonal break-up of the southern ice arch, compared to today, 790 associated with a longer open water season, caused enhanced pelagic productivity in Nares Strait and seasonal coastal melt in CMI and around NE Greenland. 4. Between 3,500 cal yrs BP and 1,400 cal yrs BP (excluding 2,500-2,100 cal yrs BP) outer Petermann Fjord was marked by enhanced pelagic productivity suggesting a marginal ice zone location during early spring/summer in response to the northern Nares Strait ice arch becoming more prominent and locally formed sea ice remaining mobile in Nares 795 Strait year-round. A stable northern ice arch throughout this interval is supported by the cessation of driftwood delivery to CMI. 5. From 2,500-2,100 cal yrs BP, declining sea-ice biomarker fluxes alongside declining productivity indicators in outer Petermann Fjord suggest a restriction of all primary productivity, likely as a result of near-perennial sea-ice cover. In combination with a small rise in the IP25 and a concomitant decrease in the HBI III fluxes in Kane Basin, strong NOW 800 conditions, reappearance of driftwood at CMI, and a local abundance peak of little auk seabird colonies at Qeqertaq (Salve Ø), this suggest a transient return to a southern ice arch regime in Nares Strait. Georgiadis et al. (2020) suggest that a dominantly positive AO phase from 3,000-1,200 cal yrs BP, associated with stronger along strait winds in Nares Strait, might have physically prevented the formation of ice arches. Since ice-arch formation in Nares Strait also depends on atmospheric temperatures and sea-ice thickness, distinct atmospheric cooling spikes between 2,500 cal 805 yrs BP and 1,900 cal yrs BP, might have exerted a positive feedback on sea-ice mechanics, counterbalancing the increased wind stress and enabling the formation of a southern ice arch.
6. The collective sea ice reconstructions in the Nares Strait and Lincoln Sea region indicate a return to a stable southern ice arch regime and potentially even double ice arching >1,400 cal yrs BP. 7. Our data demonstrates that the formation of landfast sea ice in Petermann Fjord and Nares Strait preceded major 810 growth events of the Petermann ice tongue. This suggests that sea ice promotes an environment favourable for ice tongue growth, either directly or indirectly via stabilizing the glacier calving front and possibly decreasing the inflow of AAW to Petermann Fjord. Conversely, a reduction in the landfast sea-ice season in Nares Strait and adjacent fjords, as observed during the last decade, will likely contribute to destabilizing local ice tongues, which might result in enhanced mass loss from the GrIS in the future. 815

Data availability
The data presented as part of this manuscript is archived in the PANGAEA database and can be accessed via the following link https://doi.pangaea.de/10.1594/PANGAEA.929918.

Author contribution
The study was designed by HD and CP with help from BR and AJ. MJ led the Petermann Expedition and collected the core 820 together with BR and AJ. HD carried out the sample processing and analyses for sea-ice biomarkers with help from MMJ. AJ carried out the processing and analyses for foraminiferal census counts. JO performed the TOC analyses. MJ, MG, and CP acquired funding for this study. HD prepared the manuscript with valuable contributions from all co-authors.