The Holocene dynamics of Ryder Glacier and ice tongue in north Greenland

The northern sector of the Greenland ice sheet is considered to be particularly susceptible to ice mass loss arising 20 from increased glacier discharge in the coming decades. However, the past extent and dynamics of outlet glaciers in this region, and hence their vulnerability to climate change, are poorly documented. In the summer of 2019, the Swedish icebreaker Oden entered the previously unchartered waters of Sherard Osborn Fjord, where Ryder Glacier drains approximately 2% of Greenland’s ice sheet into the Lincoln Sea. Here we reconstruct the Holocene dynamics of Ryder Glacier and its ice tongue by combining radiocarbon dating with sedimentary facies analyses along a 45 km transect of 25 marine sediment cores collected between the modern ice tongue margin and the mouth of the fjord. The results illustrate that Ryder Glacier retreated from a grounded position at the fjord mouth during the Early Holocene (>10.7 ± 0.4 cal ka BP) and receded more than 120 km to the end of Sherard Osborn Fjord by the Middle Holocene (6.3 ± 0.3 cal ka BP), likely becoming completely land-based. A re-advance of Ryder Glacier occurred in the Late Holocene, becoming marine-based around 3.9 ± 0.4 cal ka BP. An ice tongue, similar in extent to its current position was established in the Late Holocene 30 (between 3.6 ± 0.4 and 2.9 ± 0.4 cal ka BP) and extended to its maximum historical position near the fjord mouth around 0.9 ± 0.3 cal ka BP. Laminated, clast-poor sediments were deposited during the entire retreat and regrowth phases, suggesting the persistence of an ice tongue that only collapsed when the glacier retreated behind a prominent topographic high at the landward end of the fjord. Sherard Osborn Fjord narrows inland, is constrained by steep-sided cliffs, contains a number of bathymetric pinning points that also shield the modern ice tongue and grounding zone from warm Atlantic waters, and has a 35 shallowing inland sub-ice topography. These features are conducive to glacier stability and can explain the persistence of https://doi.org/10.5194/tc-2021-95 Preprint. Discussion started: 31 March 2021 c © Author(s) 2021. CC BY 4.0 License.

The sill depths on the outer fjord are 475 m on the east and 375 m on the west (Fig. 2). The inner sill has a ~6.2 km wide central region that ranges in depth from 193 to 300 m, with a ~1 km wide channel on the eastern side that extends to a depth of 390 m. The modern ice tongue terminus of Ryder Glacier is located approximately 5 km landward of the inner sill.

95
Despite these bathymetric barriers, waters of Atlantic origin that circulate through the Lincoln Sea are found between the two sills at depths greater than 350 m ( Fig. 3; Jakobsson et al., 2020). These relatively warm (>0.3 o C) and saline (>34.7) waters are constrained by the inner sill with only a small amount of warmer water flowing across the inner sill through the <1 km wide and 390 m deep channel, to be strongly mixed with glacially derived meltwater. Sherard Osborn Fjord cuts into the Paleozoic Franklinian Basin that extends over 2000 km from the Canadian Arctic to eastern north Greenland (Henriksen and Higgins, 2000). Along the north Greenland coast, deposition in the Franklinian Basin occurred in a passive margin environment through the latest Precambrian to Devonian, with shallow water carbonate shelf sediments found to the south and siliciclastic slope and deep-water sediments in the north (Henriksen et al., 2009).

105
Exposed sedimentary bedrock between the present GrIS margin and the terminus of the floating ice tongue of Ryder Glacier is composed of Cambrian and Silurian carbonate shelf sediments. Further seaward, Silurian sands and siltstones deposited in a deeper water turbiditic environment are found on the lowlands of Nyeboe and Wulff Land (Henriksen and Higgins, 2000).
Older, deformed Cambrian deep-water calcareous mudstones and sandy turbidites cap the outer headlands of Nyeboe and Wulff Land, and are part of the 600 km long E-W trending North Greenland fold belt that formed near the end-Devonian 110 Ellesmerian Orogeny (Higgins et al., 1998) (Fig. 2). Terrestrial mapping of glacial landforms and marine limits across Nyeboe Land, Warming Land and Wulff Land has defined three regional glacial events from the end of the Pleistocene (Late Weichselian) through the Holocene (Kelly and Bennike, 1992). The oldest marginal ice limits that are traced across north Greenland belong to the Kap Fulford Stade (Kelly and 115 Bennike, 1992). During this time, the GrIS extended across the southern parts of many peninsulas, with outlet glaciers flowing to the outer limits of many of the fjords. Radiocarbon dates from marine macrofossils sampled by Kelly and Bennike at Kap Fulford (Station 31;10030 ± 175 and 9390 ± 90 14 C a BP) and freshwater algae from northern Wulff Land (Station 43; 10 480 ± 90 14 C a BP) provide minimum age estimates for the retreat of glacial ice of the Kap Fulford Stade to the latest Pleistocene to Early Holocene (>10.5 cal ka BP) (Kelly and Bennike, 1992) (Table 1, Fig. 2). 120 Table 1. Key radiocarbon dates originally used to constrain the ages of glacial stages in the vicinity of Sherard Osborn Fjord. Kelly and Bennike (1992)  A second set of ice marginal deposits are found along many of the fjords, 20-60 km inland from the Kap Fulford Stade.

125
These are assigned to the Warming Land Stade, representing a standstill of outlet glaciers draining the retreating ice front. 7 The Steensby Stade marks the most recent re-advance of the ice margin, outlet glaciers and local ice caps to their maximum 140 positions since the Kap Fulford Stade. Its onset is poorly dated and occurred sometime during neoglacial cooling that extent following the Holocene Thermal Maximum. At the GrIS margin to the north of Ryder Glacier, peat deposits over which the ice margin advanced provide an age of 5100±130 14 C a BP (Station 41), while at Steensby Glacier, reworked marine macrofossils in lateral moraines yield an age of 4870±80 14 C a BP (Station 34; Kelly and Bennike, 1992). Despite uncertainty surrounding the onset of the Steensby Stade, and subsequent dynamics of Ryder Glacier through the Late 145 Holocene, the maximum extent is believed to coincide with historical observations that place the terminus of outlet glaciers and ice tongues near the mouth of many of the major fjords. This is true for Ryder's ice tongue, which was positioned near the outer margin of Sherard Osborn Fjord in 1917 by the Danish geologist and explorer Lauge Koch (Koch, 1928) (Fig. 2).

Marine sediment cores
Sediment cores from Sherard Osborn Fjord were collected during the 37-day Ryder 2019 expedition (August 5-September 10, 2019). Between August 13 and 25, unusually light sea-ice conditions allowed Oden to occupy nine coring stations and systematically map the fjord ( Fig. 2; Jakobsson et al., 2020). Coring was conducted using a 12 m long piston core (PC) and 155 small (1-2 m) trigger weight core (TWC), and a 6 m long gravity core (GC). Piston and gravity cores were collected in liners with an inner/outer diameter of 100/110 mm, while TWCs were collected in narrower (80/88mm) liners. Here we focus on cores obtained from five stations that were positioned on topographic highs, with a two radiocarbon dates coming from a sixth station on the outer sill. The five main stations form a 45 km long transect extending from the edge of the modern ice tongue to the outer sill of Sherard Osborn Fjord (Table 2, Fig. 2). Sub-bottom profiles were acquired across the coring sites with Oden's Kongsberg SBP120 (3°x3°) chirp sonar using a 2.5-7 kHz pulse. These profiles show that a 165 relatively thin (<10-15 m) drape of sediments exists on top of the acoustic basement (Fig. 4). This basement surface could either be sedimentary bedrock, or highly lithified sedimentary units like till. Hogan et al. (2020) illustrated that in Petermann Fjord, lower frequency air gun seismic data was required to differentiate bedrock from more lithified sediments and till.

170
The unsplit sediment cores were allowed to equilibrate to room temperature (~20 o C) and logged shipboard on a Geotek multi-sensor core logger (MSCL). The MSCL provided downcore measurements (1 cm sampling interval) of bulk density (using gamma-ray attenuation) and magnetic susceptibility. Bulk density is controlled by sediment porosity and grain composition (grain density). In predominantly lithogenic normally consolidated marine sediments, variability in bulk density is a function of grain size-driven changes to porosity (coarser sediments = lower porosity = higher bulk density). Magnetic 175 susceptibility is a useful proxy for the input of siliciclastic versus biogenic/organic sediments, and in dominantly lithogenic sediments can be used to discriminate between provenance and grain size (Hatfield et al., 2017;Reilly et al., 2019). Gammaray attenuation was measured using a 137 Cs source with a 5 mm collimator and a 10 s count time. applied to account for the difference between the core diameter (D C ) and the loop diameter (D L ). The volume-specific 180 magnetic susceptibility (κ, 10 -5 SI) is defined as, The mass-specific magnetic susceptibility (χ) was calculated by dividing κ by the bulk density.

185
After logging, piston and gravity cores were split into working and archive half sections. Archive halves were described and imaged using a digital line-scanning camera on the MSCL, while the working half was sampled for shore-based analyses.
Split sections were wrapped in plastic film, placed in D-tubes and stored in a refrigerated container (4 o C) until they were returned to Stockholm University's refrigerated core storage facility. TWC's remained unsplit and placed directly into the 190 refrigerated container.
Following the expedition, all archive halves were analysed on an Itrax X-ray fluorescence (XRF) core scanner at the B60s (sharp). Images were processed using SedCT MATLAB tools to obtain a coronal slice through the central region of each core at an effective pixel resolution of 0.5 x 0.5 mm . Clasts with >1 mm diameter were automatically counted from the 3-dimensional CT data, binned into 2 cm intervals, and normalized by core volume using the algorithm of Reilly et al. (2019) which provides an objective proxy for ice rafted debris (IRD) concentration in glaciomarine lithofacies deposited beyond the limit of grounded ice.

205
Twenty-nine grain size measurements were made on core 6-GC to broadly characterize grain size spectra in the different lithologic units. Measurements were performed using a Malvern Mastersizer 3000 laser diffraction particle size analyzer at

Composite depth scales
The first attempts at coring in Sherard Osborn Fjord illustrated that near surface sediments were very soft, and the top of the coring tools were sinking below the seafloor resulting in variable the recovery of near surface sediments between the gravity, piston and trigger weight cores. At stations 7, 8 and 9, multiple coring tools were deployed, and sediments were recovered 215 using all of these devices. A composite depth scale was developed for each station to ensure that measurements from the different cores could be integrated. The composite depth scales were developed through correlation of the MSCL bulk density, magnetic susceptibility and XRF-scanning (mainly Ca content) data and additionally through visual correlation using the CT-images (supplementary material Table S1).

220
The piston core at each station was used as an undistorted reference depth scale. Depths of the gravity and trigger weight cores were stretched, or compressed, between tie points to the piston cores. At stations 7 and 9, the piston core recovered sediments closer to the seafloor than the gravity or trigger weight cores. At station 8, correlation of MSCL and XRF data indicate that the gravity core recovered 40 cm of near-surface sediments that were either not recovered or lost from the top of the piston core.

Radiocarbon dating
Radiocarbon measurements ( 14 C) were made on 48 samples (

235
There is no a priori information on the local marine reservoir correction (ΔR) for Sherard Osborn Fjord or the Lincoln Sea.
Estimates of ΔR for nearby regions vary widely and have been derived using different marine calibration curves. Results from three pre-bomb living molluscs collected in the vicinity of Thule, suggest a ΔR of 5 +/-50 years (Mörner and Funder, 1990). Kelly and Bennike (1992) applied 150-year ΔR for marine macrofossils from the region surrounding Ryder Glacier, as suggested for areas of northernmost Greenland by Funder et al. (1982). However, Coulthard et al. (2010) found an 240 average ΔR of 335±85 years (using Marine09; Reimer et al., 2009) based on 24 molluscs from the northwestern Canadian Arctic Archipelago, which includes the northern and western coasts of Ellesmere Island. Reilly et al. (2019) argued that a ΔR of 770 years (using Marine13; Reimer et al., 2013) provided the best fit between a stacked paleosecular variation record from Petermann Fjord sediments and a North Atlantic reference curve. This large ΔR was also consistent with large offsets between radiocarbon dates on planktic and benthic foraminifera and 210 Pb derived chronologies for the past 100 years. In 245 summary, estimates used in the literature range from 0-770 years. In this study we have applied the new Marine20 calibration curve (Heaton et al., 2020), which results in ages that are ~150 years younger than equivalent Holocene 14 C ages calibrated using Marine09 or Marine13. Taking this into consideration we have applied a ΔR of 300±300 years. This broad range provides a large uncertainty envelope. The upper bound approaches the older offset found by Reilly et al. (2019) for Petermann Fjord (770 years using Marine13, which equates to ~620 years using Marine20) and the lower bound the 150 250 years (which equates to ~0 years using Marine20) commonly used to reconstruct terrestrial ice margins in the area (Kelly and Bennike 1992;Young and Briner, 2015). The mean calibrated ages obtained using a ΔR of 300±300 years (equivalent to ~450 years using Marine13) provide a suitable estimate for sites that are influenced by Atlantic waters. This dependency on water mass is one of the underlying problems in determining an applicable local reservoir correction. For example, the paired benthic/planktic foraminifera samples we ran revealed offsets of 470-570 years in 6-GC and 7-PC which are currently bathed 255 in Atlantic waters, and 0 to 200 years at 10-GC which lies closer to the mixed surface layer ( Fig. 3; Table 3). Additional work is needed to resolve the issue of local marine reservoir offsets in the region, their dependency on water mass, and how they have may have changed through time.

Lithostratigraphic Units
Data from the MSCL, XRF-scanning data, and CT imaging are used to identify six major lithologic units that are correlated 265 from the fjord mouth (10-GC) to the inner bathymetric sill lying seaward of the modern ice tongue (coress 8-PC and 9-PC) ( Fig. 5). A sixth core, 11-GC, located on the outer sill in the shallowest water depth (208 m), did not contain the same lithostratigraphic units, and only the basal radiocarbon date is used as a constraint for Early Holocene glacial ice retreat.

LU6:
The lowermost lithologic unit (LU6) was recovered in 7-PC, 8-PC and 9-PC. It has moderately high but variable Ca 270 contents and is a coarse-grained, poorly sorted diamict containing abundant gravel (2-64 mm) through cobble-sized (64-256 mm) clasts with a high bulk density (> 2 g/cm 3 ) (Figs. 5, 6). The mass-specific magnetic susceptibility is considerably higher in LU6 compared to the overlying LU5, generally exceeding 80-100 x10 -8 m 3 /kg. Less than 5 cm of this unit were recovered in 9-PC, with greater recovery in 7-PC (54 cm) and 8-PC (>135 cm). The upper boundary with LU5 is abrupt in 8-PC, gradual in 7-PC. In 9-PC, the thin LU6 and lowermost LU5 sediments are laminated, deformed and contain two large 275 dropstones, making it difficult to determine the nature of the boundary. The LU6/LU5 boundary is marked by a pronounced up-core decrease in bulk density and mass-specific magnetic susceptibility. Two subunits (LU6a and LU6b) are recognized in the longer sequences from 7-PC and 8-PC. These subunits are separated by a sharp boundary (Fig. 6) are sometimes visible in the split core sections as alternating reddish-brown and lighter tan-colored laminae (Fig. 6). The laminations are best defined by variations in Ca and bulk density. Ca enrichment arises from increased inputs of detrital carbonate, eroded from the surrounding Cambrian and Silurian bedrock. Where visible to the naked eye, the lighter tancolored laminations are enriched in Ca and denser (lighter in CT-scanning images) (Fig. 5). Ca-enriched layers are skewed towards medium and coarse silts, while laminae with lower Ca abundance are skewed towards fine silt and clay (Fig. 7). The 290 higher bulk density of the light-colored laminae likely is caused by the coarser grain size and higher detrital carbonate content, which has a higher grain density compared to quartz and clay minerals. LU5 has a low abundance of irregularly spaced clasts >1 mm that are more prevalent near the base of the unit (Fig. 5). In cores 10-GC, 7-PC, and 6-GC, there is a notable up-core decrease in the bulk density, Ca-content and laminae thickness of LU5. Mass-specific magnetic susceptibility also increases up-core, as the relative Ca-abundance decreases, consistent with a larger contribution from 295 siliciclastic sediments. In 6-GC and 7-PC, these long-term trends are interrupted by a 0.50-1.0 m thick interval with elevated Ca contents and thicker laminae, found near the top of the unit.
CT images illustrate that laminations in LU5 are wavy to lenticular in nature, and hence deposited in part by traction processes under the influence of bottom water currents (Fig. 6). Throughout LU5, the laminations are disrupted by numerous 300 small-scale normal faults. In 6-GC and 7-PC, faulting in LU5 is more pronounced than in the overlying laminated sequence of LU4 (Fig. 6). Laminations become less convoluted and more planer towards the top of LU5, indicating an overall reduction in traction dominated sedimentation through time.
At the two stations from the inner sill, 8-PC (238 mbsl) and 9-PC (274 mbsl), LU5 is considerably thinner than in 6-GC (633 305 mbsl) and 7-PC (559 mbsl), which lie further seaward, and in deeper water depths (Fig. 5). The thinner sequences of LU5 in 8-PC and 9-PC also contain intervals where the laminations appear to be truncated or eroded, sometimes with a lag deposit of coarse-grained material (see 9-PC-6 in Fig. 6), indicating that deposition of LU5 was discontinuous at these shallower water stations. On the outer sill, LU5 is also comparatively thin, containing highly fractured and wispy laminations that are truncated or eroded in some intervals. This also indicates a more dynamic depositional environment, likely influenced by 310 current activity or ice scouring. The base of LU5 in 10-GC is dominantly composed of sand and gravel sized material.
LU4: Faint, planar mm-scale laminations that lack notable evidence for bioturbation define LU4. The transition between LU5 and LU4 is gradual, with the base of LU4 being identified by a notable and correlative decline in Ca-abundance that is identified at all the coring stations (Fig. 5). Laminations in LU4 are only weakly visible by the naked eye but are evident in 315 the CT scanning data (Fig. 6). In 7-PC and 6-GC, laminations appear to transition from wavy to lenticular with frequent faulting in LU5, to planar in LU4, indicating an increased influence of suspension settling on sedimentation. With the exception of 10-GC on the outer sill, coarse ice-rafted clasts remain dispersed in low numbers through LU4. LU4 is a silty clay, containing very minor amounts of sand (<1%), with some samples showing a slight elevation of medium to coarse silt (Fig. 7). The bulk density remains relatively constant (1.6-1.8 g/cm 3 ) with mass specific magnetic susceptibility higher than 320 in LU5, and increasing up-core towards the top of the unit, consistent with a continued up-core decrease in detrital carbonate concentrations. Occasional minor pulses of Ca are seen throughout the unit. The LU5 to LU4 transition appears correlative across the fjord, and many small-scale features in the Ca, bulk density and magnetic susceptibility logs can be traced between the sites (Fig. 5). This suggests a uniform depositional environment across the fjord during this time interval.
LU3: Across Sherard Osborn Fjord, LU4 is capped by LU3, a second diamict containing large gravel to cobble-sized clasts (Figs. 5,6). LU3 is recognized by its high bulk density (1.8-2.0 g/cm 3 ), elevated Ca-content and abundant clast content (10 ->100 clasts/cm 3 ) (Fig. 5). LU3 has a low mass specific magnetic susceptibility, except at 8-PC where cobble and bolder sized clasts of felsic rocks were recovered, which presumably led to the high susceptibility measurements. CT images from 6-GC and 7-PC show that the lower boundary is heavily bioturbated, while at 8-PC and 9-PC it is marked by a higher 330 concentration of coarse clasts that tend to fine upwards (Fig. 6).  5), is composed primarily of fine silt and clay, and contains very few scattered ice-rafted clasts (Fig. 6). The thickness of LU2 varies considerably across the fjord, with the thickest occurrences in 6-GC and 7-PC. LU2 is nearly absent in sediments from the outer sill (10-GC) (Fig. 5).

345
LU1: A return to a laminated facies lacking coarse ice-rafted clasts and bioturbation defines LU1. The grain size spectra of LU1 closely resemble those from LU4 (Fig. 7). The laminations in LU1 are also highly fractured and distorted along the sides of the core. This is attributed to coring disturbance and the soft to soupy nature of these near surface sediments. LU1 is divided into two subunits. LU1b is only found at stations 8 and 9 on the inner sill. It is a faintly laminated unit, with some discrete intervals where the laminations are disturbed, sometimes by bioturbation. LU1b has low abundances of Ca, similar 350 to what is found in LU2 and LU4 (Fig. 5). LU1a is marked by the emergence of more pronounced mm-to cm-scale laminations that are only seen clearly in the CT images (Fig. 6). LU1a is marked by increased Ca concentrations, which are most apparent in 9-PC and 7-PC (Fig. 4). At stations 10, 6 and 7, LU1a sits abruptly on top of LU2 (Fig. 5), while at stations 8 and 9, there is a gradual transition between LU1b and LU1a.

Ages of unit boundaries
Radiocarbon dating is used to constrain the ages of lithostratigraphic unit and subunit boundaries. In most cases samples 365 were collected from within a few centimeters of these boundaries in one or more cores (Fig. 5). The youngest age underlying a unit or subunit boundary is used to date the boundary (Fig. 8). However, this is not possible for the LU5/LU6 and LU1b/LU2 boundaries. For LU6 the oldest age from the overlying unit is adopted as the boundary age. A younger than and older than age is provided for LU1b, based upon dates obtained from two different cores (Table 4). Of the 48 radiocarbon dates, 7 were identified as outliers because the overlying ages are older and fell outside the 1σ calibrated age range. The 370 lowermost sample in 8-PC (Sample #35, Table 3) is also considered an outlier, due to the substantially younger age returned in a sample 22.5 cm up-core (Sample #34, Table 3). LU5 sediments from this site were not deposited continuously, implying periods of erosion or non-deposition that could account for the large difference in ages between the lowermost dates in 8-PC. The lithologic sequence from Sherard Osborn Fjord sediments spans nearly the entire Holocene (Table 4), with the oldest 375 boundary (LU5/LU6) dated in 7-PC (Sample #24) to 10200±40 14 C years (10330-11160 cal. a BP) while the youngest boundary (LU1a/2) is dated in 6-GC (Sample #1) to 1730±40 14 C years (540-1150 cal. a BP). The oldest age comes from 11-GC (Sample #48, 11150±250 14 C years) and provides a constraint on the timing for ice retreat from the outer sill.  (Table 4). On the inner sill, the age range for the LU2-LU1b transition is bracketed by an 'older than' and 'less than' age from 9-PC and 8-PC respectively. Index numbers correspond to those in Tables 3 and 4   410 (Koch, 1928) (Fig. 2).

Early Holocene glacier and ice tongue dynamics
The oldest recovered sediments come from 11-GC on the outer sill where a sample of the benthic foraminifera Cassidulina teretis returned a date of 12±0.5 ka. Due to the limited amount of sediment that has accumulated above the bedrock/till at 6-GC, 10-GC and 11-GC (Fig. 4), we interpret this as a minimum age for the retreat of Ryder Glacier from the outer sill, where it was grounded 80-90 km seaward from its current position (Fig. 10). Deposition of the grounding zone proximal meltwater 425 sediments (LU5) above subglacial diamicts in 7-PC commenced at 10.7±0.4 cal ka BP, indicating that the glacier had retreated further inland by this time (Fig. 10). The calibrated ranges of radiocarbon dates from the base of 7-PC (10.3-11.1 cal ka BP) conform to the reported age for the Kap Fulford Stade (>10.5 cal ka BP) (Kelly and Bennike, 1992) (Fig. 10). It is less certain if the older age from 11-GC on the outer sill (11.6-12.6 cal ka BP) (Fig. 8,  (Fig. 2). The strong meltwater signal captured in the LU5 coincides with the rapid Early Holocene reduction in the height of the Greenland ice sheet (Lecavalier et al., 2017) (Fig. 9).

435
Sediments from LU5 are comparatively thin at the inner sill coring sites, and difficult to directly correlate with the records from 6-GC and 7-PC. Erosional events identified in LU5 sediments from the inner sill suggest discontinuous sedimentation (see 9PC-6 Fig. 6). Sitting in much shallower water depths (228-271 mbsl; Table 1), glacial ice likely remained grounded here after it retreated from stations 6 and 7. Deposition on the inner sill would also have been influenced by bottom scouring 440 currents and debris flows while the glacier was grounded (or near flotation) on shallow regions of the sill that rise to depths of 193 m. Therefore, we suggest that Ryder Glacier remained grounded on the inner sill until sometime after 9.3±0.4 cal ka BP, which marks the onset of continuous sedimentation of LU5 at this location ( Fig. 10; Table 4). This is consistent with the age of the Warming Land Stade, bracketed between >9.5 cal ka BP and 8.0 cal ka BP by Kelly and Bennike (1992). The shallow inner sill, first identified through bathymetric mapping during the Ryder 2019 expedition , is 445 very close to the previously inferred grounding zone location for the Warming Land Stade and provides a natural pinning point for the marine-based glacier during this standstill (Fig. 2).
In 6-GC and 7-PC there is a broad period of elevated detrital carbonate delivery (high Ca-content) that occurs near the top of LU5 (Fig 5). This likely reflects enhanced meltwater delivery during terrestrial ice retreat at the end of the Warming Land 450 Stade prior to the LU4/LU5 boundary at 7.6±0.40 cal ka BP ( Fig. 10; Table 4). Across Sherard Osborn Fjord, deposition of the laminated LU4 sediments continued until 6.3±0.3 cal ka BP ( Fig. 10; Table 4), which is very similar to the age (6.0 cal ka BP) for the minimum Holocene extent reported for Ryder Glacier (Kelly and Bennike, 1992).

Middle Holocene inland retreat and collapse of Ryder's ice tongue
The second diamict facies (LU3) in the stratigraphic succession separates the faintly laminated sub-ice tongue sediments of LU4 and the bioturbated facies of LU2. As this unit is found from the inner sill all the way to the outer sill, and does not appear to be time-transgressive, it cannot be attributed to a slow landward migration of the ice tongue calving margin.
Instead, it is interpreted as a collapse facies (Smith et al., 2019) associated with the sudden disintegration of Ryder's ice 465 tongue. While this interpretation fits with the overall facies succession, the ~2.4 kyr duration of LU3 (from 6.3±0.3 to 3.9 ± 0.4 cal ka BP; Table 4) is too long to be a simple abrupt collapse event, as these are generally associated with increased sediment accumulation rates (Smith et al., 2019).
An alternate interpretation is that LU3 sediments represent an erosional event. However, this does not fit with the 470 overlapping ages of sediments from the top of LU4 at coring stations 6, 7, 9 and 10, and to a large degree the ascending ages found downcore through LU3 at station 8 (Fig. 8). Therefore, the widespread distribution, long duration, consistent boundary ages of LU3, and lack of evidence for pervasive ice scouring makes it hard to reconcile this facies with a glacial surge. This unit is also much younger than the inferred age for the drainage of the ice dammed lake that developed on Wulff Land between Ryder and C. H. Ostenfeld glaciers between 7.9-8.5 cal ka BP (Kelly and Bennike, 1992). These observations 475 suggest that LU3 marks a period of low sedimentation that began around 6.3±0.3 cal ka BP when terrestrial studies suggest that Ryder Glacier had retreated further inland than its current position (Kelly and Bennike, 1992).
The explanation that best fits evidence from terrestrial field studies, and the overall facies succession, is that the condensed diamict of LU3 was deposited when Ryder Glacier retreated far enough inland to become cut-off from the main fjord. In 480 Sherard Osborn Fjord, a relatively deep, isolated marine embayment exists behind a prominent topographic high lying 40 km inland of the modern grounding zone (Fig. 11). Here elevations increase to between 100-200 m above sea level, with elevations surrounding the embayment reaching 400 -600 m above sea level (Figs. 10, 11). Retreat of the grounding zone onto this topographic high (possibly a former grounding zone wedge), or even inland of it, would explain the apparent rapid disintegration of the ice shelf at the end of LU4. The entrainment of meltwater-derived sediments within the isolated 485 embayment or even behind the inner sill, can account for the overall reduction in sedimentation rates in the outer fjord. This is consistent with observations from modern proglacial lakes that illustrate their efficiency at disrupting meltwater fluxes and sequestering sediments (Carrivick and Tweed, 2013;Bogen et al., 2015;Piret et al., 2021). While grounded in this distal position, calving could have discharged some icebergs into Sherard Osborn Fjord, delivering coarse-grained ice rafted sediments. Elevated Ca contents within LU3 (Fig. 5) attest to sediment provenance being primarily from landward of the 490 inner sill (Fig. 2). Furthermore, the sedimentation rates during LU3 are low (9-10 cm/kyr; Fig. 9) and comparable with sedimentation rates of 10-30 cm/kyr reported by Dowdeswell et al. (1994)  fjord or beyond is the lack of iceberg scouring seen in bathymetric data on the inner sill. This suggests that the calving front was grounded in water shallow enough to not produce icebergs in excess of 190-200 m thickness -a condition that can 495 easily be met by the landward retreat of Ryder Glacier (Fig. 10).
A modern analogue for this configuration is seen in neighboring Victoria Fjord. Today C. H. Ostenfeld Glacier is grounded on a topographic high at the end of the fjord, in an equivalent position to the landward terminus of Sherard Osborn Fjord where the proposed Middle Holocene ice margin is located (Fig. 10). Between 2000 and 2006, the C. H. Ostenfeld ice 500 tongue, which protruded from a narrow over-deepened marine channel, largely disintegrated (Moon and Joughin, 2008), while the grounding zone may soon become increasingly land-based and discharge grounded ice directly into Victoria Fjord (Hill et al., 2017) (Fig. 10).
The continued retreat of Ryder Glacier through the Early and Middle Holocene is consistent with many regional 505 paleoclimate proxies from around northern Greenland. This includes beach ridge formation along the north Greenland coast, indicating more open water conditions that persisted until ~6 cal ka BP (Möller et al., 2010;Funder et al., 2011), reduced winter sea-ice conditions in Bliss Lake until 6.5 cal ka BP (Olsen et al., 2012), and peak late summer air temperatures inferred from δ 18 O of chironomids in Secret and Deltasø lakes that were 2.5-4 o C warmer then present until 6.2-6.0 cal ka BP (Axford et al., 2019;Lasher et al., 2017) (Fig. 1). The paleotemperature estimates from these lakes indicate a slight cooling, 510 but persistent, stable and positive temperatures between ~6.0-4.0 cal ka BP (Lasher et al., 2017;McFarlin et al., 2018).
Diminished sediment delivery due to the presumably stable inland position of Ryder Glacier lasted from 6.3±0.3 to 3.9±0.4 cal ka BP (Table 4). The inferred re-advance into Sherard Osborn Fjord, marked by increased sediment delivery at the onset of LU2, occurred around 3.9±0.4 cal ka BP. This timing for glacier advance is consistent with cooling seen in lake based temperature reconstructions around 4 cal ka BP (Lasher et al., 2017) and the oldest estimated age (3.5 to 4.0 cal ka BP) for 515 ice at the base of the southern dome of Hans Tausen ice cap, which had disappeared during the Middle Holocene (Madsen and Thorsteinsson, 2001;Landvik et al., 2001;Zekollari et al., 2017).
A stable inland position for Ryder Glacier during LU3 fits with generally colder conditions that prevailed over north Greenland towards the end of the middle Holocene. This is reflected in the development of more prolonged winter ice cover 520 in Bliss Lake after 5.9 cal ka BP (Olsen et al., 2012) and a period of low melt rates for the north Greenland ice sheet inferred from elevation changes at Camp Century (Lecavalier et al., 2017) (Fig. 9). Therefore the prolonged slow deposition of coarse IRD during LU3 is likely a combined effect of limited but persistent iceberg calving, the slow melt-out of englacial material from the disintegrated ice tongue, as well as continued input of sea-ice rafted material entrained during periods of shore-fast sea ice growth and decay.

Late Holocene cooling and the re-growth of Ryder's ice tongue
The advance of Ryder Glacier and the re-growth of local ice caps surrounding Sherard Osborn Fjord is described by the 535 Steensby Stade re-advance. Its onset is not well constrained by terrestrial field data, but is believed to have started sometime after 5.1-4.7 ka (Kelly and Bennike, 1992), with outlet glaciers reaching their maximum extents by the start of the 1900's.
Thus, in marine records from Sherard Osborn Fjord, the Steensby Stade encompasses deposition of LUs 2 and 1.
The highly bioturbated LU2 sediments appear abruptly on top of LU3, with the onset of deposition sometime after 3.9 ± 0.4 540 cal ka BP (Table 4). While the inland glacier may have started advancing earlier, increased sedimentation rates at the onset of LU2 (Fig. 9) likely marks the seaward migration of Ryder Glacier to a marine-based position in the innermost fjord.
Within this setting, the re-growth of a limited ice tongue would act as a filter for coarse iceberg rafted sediments and can explain the paucity of IRD in LU2 sediments. Extensive bioturbation in LU2 attests to warmer climatic conditions and the absence of an ice tongue seaward of the mapped inner sill and likely less persistent sea-ice cover.

545
On the inner sill, deposition of laminated sediments lacking bioturbation (LU1b) marks the re-growth of the ice tongue out to this location (Fig. 10) between 3.6 ± 0.4 and 2.9 ± 0.4 cal ka BP. This is roughly 300-1000 years after the glacier became marine-based at the end of LU3. The persistence of bioturbation (LU2) at stations in the middle fjord (stations 6 and 7), indicates an earlier establishment of the ice tongue over the inner sill, while more open water conditions persisted further 550 seaward (Fig. 9). This configuration is similar to what we found during our expedition in 2019 (Fig. 2). It appears that this was a relatively stable configuration that persisted for another 2-2.7 kyrs (until 0.9 ± 0.3 cal ka BP; Table 4), before the ice shelf extended towards the outer sill and laminated LU1a sediments were deposited in the middle fjord (Figs. 9, 10). The long-term stability of the ice tongue terminus near the inner sill is consistent with the inner sill acting as a barrier for Atlantic water invasion and limiting basal ice tongue melting .

555
The Late Holocene re-growth of Ryder's ice tongue has some parallels with the Petermann ice tongue, which began to reform between 1.9-2.3 cal ka BP, after being absent since its collapse around 6.9 cal ka BP (Reilly et al., 2019). Petermann also attained a stable ice tongue with an extent similar to 20 th century historical observations between 0.4-0.9 cal ka BP (Reilly et al., 2019). At both Ryder and Petermann the growth of ice tongues towards the outer fjords occurred much later 560 than the establishment of multi-year landfast sea ice in front of Phillips Inlet and Disraeli Fjord on northern Ellesmere Island around 5.5 cal ka BP (England et al., 2008). On the other hand, ice tongue re-growth in Petermann (1.9-2.3 cal ka BP) and growth of Ryder's ice tongue to the inner sill in Sherard Osborn Fjord (by 2.9 ± 0.4 cal ka BP) are consistent with the development of more extensive sea ice around the northern Greenland margin by 2.5 cal ka BP (Funder et al., 2001) and only intermittent periods of sea-ice free conditions in the Lincoln Sea after 3.9 cal ka BP based on the cessation of driftwood 565 delivery to Clements Markham Inlet (England et al., 2008) (Fig. 9).
Uncertainty in the local reservoir correction leaves the timing for the maximum Late Holocene growth of Ryder's ice tongue poorly constrained. However, as with the final growth of Petermann's ice tongue (Reilly et al., 2019), it appears consistent with numerous regional paleoclimate records that document the development of cooler conditions close to the transition from 570 the Medieval Warm Period to the Little Ice Age. For example, geochemical proxies from Bliss Lake (Fig. 1) indicate the onset of cooling around 850 cal. a BP (Olsen et al., 2012), which agrees with the onset of cooler air temperatures inferred from the δ 18 O record of the Hans Tausen ice cap around 780 cal. a BP (Olsen et al., 2012;Hammer et al., 2001). The absence of a thin bioturbated facies reflecting deposition since the 20 th century retreat of Ryder's ice tongue is likely due 575 to incomplete recovery of surface sediments in the studied cores. Given the predicted range of sedimentation rates for LU1a in 7PC (51-102 cm/kyr, Fig. 9), only 4-8 cm of sediment would have accumulated since the retreat of Ryder's ice tongue to its current position ~80 years ago. This thin sedimentary cover could easily have been missed during coring, especially considering the soft nature of surface sediments encountered in the fjord.

580
The retreat of Ryder Glacier from the fjord mouth (>10.7±0.4 cal ka BP) is consistent with the demise of glacial ice across much of northern Greenland in the Early Holocene (Bennike and Björck, 2002;Larsen et al., 2010;Strunk et al., 2018;Larsen et al., 2018). During the early stages of deglaciation, the Lincoln Sea was covered by thick shelf ice, fed by coalescent glaciers emanating from Greenland and Ellesmere Island (England et al., 1999;Larsen et al., 2010).  (Jennings et al., 2011(Jennings et al., , 2019Georgiadis et al., 2018). Therefore, 600 differences in the timing of retreat between Petermann and Ryder Glaciers appears closely tied to local glacial and sea-ice conditions. These appear to have been less severe and more mobile in the Lincoln Sea in the Early Holocene compared to southern Nares Strait region.
The sedimentary facies from Sherard Osborn Fjord documents seemingly gradual retreat and regrowth phases for Ryder 605 Glacier as evidenced by the generally low abundance of coarse IRD (Fig. 9)  ka BP is marked in sedimentary records by the abrupt appearance of IRD clasts in sediments from across the fjord (Reilly et al., 2019). The stability of Ryder Glacier's ice tongue is most likely related to the physiography of Sherard Osborn Fjord.

610
The inner sill in Sherard Osborn Fjord not only acts as a natural pinning point for glacial ice during its retreat but also impedes the flow of warm Atlantic subsurface waters into the innermost fjord ( Fig. 3; Jakobsson et al., 2020). This likely reduces the subaqueous melting at the grounding zone and of the ice tongue today, and may have done so efficiently for much of the Holocene. The subglacial bed topography and physiography of the inner fjord will also exert an influence on glacier (and hence ice tongue) stability. There appears to be a slight retrograde slope between the inner sill and modern 615 grounding zone (Fig. 10), which are common in northern Greenland fjords and can be conducive to rapid glacial retreat (Enderlin et al., 2013;Carr et al., 2015;Hill et al., 2018). However, Sherard Osborn Fjord also narrows considerably landward of the inner sill and remains flanked by steep-sided cliffs (Fig. 11). Jamieson et al. (2012) have shown that even on a retrograde slope, retreating ice streams slow or can become pinned where they pass narrow sections in surrounding topography. The sedimentary record from seaward of the inner sill contains no evidence for massive calving events that 620 would otherwise suggest a collapse of the ice tongue during a rapid retreat down a retrograde slope. However, there is tremendous uncertainty regarding the bathymetry beneath the current ice tongue as no direct measurements exist, so even the presence of a retrograde slope is not certain. A more certain feature of the bathymetry is the steeply rising bed topography landward of the modern grounding zone (Figs. 10, 11). This feature, which could be a former grounding zone, may have promoted the gradual retreat of Ryder Glacier and the stability of its ice tongue during the Early and Middle Holocene 625 (Powell, 1990;Alley, 1991;Hill et al., 2018). A critical observation is that despite the apparent persistence of Ryder's ice tongue, the physiographic setting of Sherard Osborn Fjord, which appears to be conducive to glacier stability, did not halt the overall retreat of Ryder Glacier under the relatively mild changes in climate forcing during the Holocene.
From the results of this study, it is not possible to constrain the amount of inland retreat Ryder Glacier experienced during 630 the Middle Holocene. Based on the cessation of sedimentation in the outer fjord, we suggest that it may have retreated at least into the restricted embayment at the end of Sherard Osborn Fjord (Figs. 10, 11). Unlike Petermann Glacier, where the bed topography widens and extends far beneath the GrIS at elevations below modern sea level, the bed of Ryder Glacier abruptly rises above sea level some 60 km landward of the modern grounding line (Fig. 11). This is comparable to the smaller glaciers in the immediate surroundings and to the Køge Bugt glacier complex in southeast Greenland, where the 635 steep land-ocean boundary limited the Middle Holocene retreat of the glaciers in this area . This suggests that the land-ocean interface might have been an important boundary that mediated or halted the Holocene retreat of many glaciers that are not located in deep troughs that extend beneath the inner ice sheet. Overall, our constrains on the Holocene dynamics of Ryder Glacier suggest that it could retreat another 40-60 km inland if climatic conditions remain similar or exceed those of the Middle Holocene.

Conclusions
Like much of northern Greenland, Ryder Glacier responded acutely to climate variability in the Holocene. During the Early and Middle Holocene it retreated over 120 kilometers from a grounded position near the mouth of Sherard Osborn Fjord (80 km seaward of the modern grounding zone) to likely become land-based more than 40-60 km landward of its current position by 6.3±0.3 cal ka BP. Throughout this long period of retreat, deposition of laminated, clast poor sediments attest to 645 strong meltwater inputs and an overall stable ice tongue. Ryder Glacier remained land-based until the Late Holocene (3.9 ± 0.4 cal ka BP). As it again advanced into Sherard Osborn Fjord, an ice tongue developed that quickly grew out to its 21 st century position near a prominent bathymetric sill, located 30 km seaward of the modern grounding zone. Today this sill reduces the incursion of Atlantic waters into the inner fjord and shields the grounding zone and ice tongue from basal melting . This was also likely the case for 2.0-2.7 kyrs in the late Holocene, before the ice tongue 650 grew to its historical maximum extent around 0.9±0.3 cal ka BP. Late Holocene bioturbated sediments that are nearly devoid of ice rafted debris were deposited between 3.9±0.4 and 0.9±0.3 cal ka BP and indicate reduced ice cover over much of the middle and outer fjord. This contrasts with the near permanent shorefast ice conditions that existed in Sherard Osborn Fjord for much of the late 20 th century (Higgins, 1989), but it is similar to conditions during the summer of 2019. The physiography of Sherard Osborn Fjord appears to have had a stabilizing effect on the ice tongue of Ryder Glacier through 655 the Holocene. Nevertheless, Ryder Glacier still retreated >40-60 km landward of its current position and remained there