Rapid Reconfiguration of the Greenland Ice Sheet Coastal Margin
Abstract
The Greenland Ice Sheet has lost mass at an accelerating rate over the last two decades, but limits of early remote sensing restricted examination of localized change at an ice-sheet-wide scale. We use satellite-derived ice sheet surface velocities, glacier terminus advance/retreat, and surface elevation change data spanning ~1985–2015 to explore local characteristics of what is now a rapid reconfiguration of the ice sheet coastal margin. Widespread glacier terminus retreat is a more consistent climate response indicator than surface velocities, though local velocity patterns provide indicators of ice flow reconfiguration, including narrowing zones of fast-flow, ice flow rerouting, and outlet abandonment. The implications of this observed rapid reconfiguration are wide ranging and likely include alteration of subglacial hydrology, iceberg discharge, liquid freshwater flux, potential nutrient and sediment flux, and mass flux. Without detailed observations of earlier deglaciations and with present limits on ice sheet model capabilities, these observational records provide an important analogue for past deglaciation and for projecting future ice loss.
Key Points
- Ice sheet surface velocity and ice edge records since the mid-1980s show rapid ice sheet reconfiguration during the 21st century
- Despite widespread retreat, glaciers sped up and slowed down, revealing upstream links and strong topographic control of ice sheet behavior
- Reconfiguration on decadal timescales includes flow rerouting, narrowing of fast flow zones, and likely outlet abandonment
Plain Language Summary
From the 1970s through the early 1990s, the Greenland Ice Sheet was roughly in balance, with mass gains equaling mass losses. In the mid-1990s, however, Greenland ice loss began and accelerated. By combining ~1985–2015 records of changing outlet glacier flow, ice edge positions, and ice sheet surface elevation, we show that the margin of the Greenland Ice Sheet is undergoing a significant reconfiguration. Ice edge retreat is ubiquitous, with virtually no glaciers experiencing advance, while some areas of the ice sheet have sped up and others have slowed. Our observations reveal a rapid reconfiguration around the full ice sheet margin, with narrowing areas of fast ice flow, changes in the routing of ice flow, and glacier outlets that are likely being abandoned. The implications of rapid ice sheet reconfiguration are wide ranging. Water movement underneath the ice sheet likely changes, along with the quantity and timing of iceberg production and freshwater input to the ocean, affecting the nutrients and sediment transport from the ice sheet to local and regional ecosystems. Without detailed observations of earlier deglaciations and with limits on ice sheet computer simulation capabilities, these observational records provide an important analogue for past deglaciation and for projecting future ice loss.
1 Onset and Drivers of Rapid Ice Sheet Reconfiguration
Changes in Greenland Ice Sheet mass and dynamics have wide-ranging impacts. Pan-Greenland and regional ice sheet mass loss influences sea level in all coastal areas around the world (Larour et al., 2017; Mitrovica et al., 2018). Increasing freshwater flux affects regional ocean properties (Luo et al., 2016; Moon et al., 2017; Straneo & Heimbach, 2013) and changing ice dynamics impact nutrient, sediment, and solid ice fluxes, and local ecosystems (Bendixen et al., 2017; Cape et al., 2018; Hendry et al., 2019). What are the indicators of contemporary Greenland Ice Sheet change? How are ice sheet-wide changes reflected on the local scale? Recent and continued advances in remote sensing, and the simple passage of time, now enable collection and analysis of the multidecadal records needed to address these questions.
Cumulated evidence reveals substantial change across the full ice sheet, on catchment to continental scales, concentrated within roughly the last two decades. Across the 20th and 21st centuries, Greenland went through a period of mass loss centered around the 1930s, followed by several decades of variation between mass balance and mass loss (Box & Colgan, 2013; Kjeldsen et al., 2015). Entering the most recent half century, the ice sheet was roughly in balance during the 1970s, 1980s, and early 1990s (note, however, that glaciers did not overall readvance to pre-1930s positions) (Fettweis et al., 2017; Mouginot et al., 2019). This relatively stable behavior collapsed in the mid 1990s, and the decades since have been marked by a rapid and accelerating increase in ice mass loss (Khan et al., 2015; Mottram et al., 2019; Mouginot et al., 2019).
The causes of ice mass loss are multiple, with increases in ice sheet surface and submarine melt across the full ice sheet margin. Ice sheet surface melt is increasing due to overlapping processes including general atmospheric warming and alterations in atmospheric circulation patterns such as the North Atlantic Oscillation, Atlantic Multidecadal Oscillation, and Greenland Blocking Index, which influence air pressure, temperature, and cloudiness (Bevis et al., 2019; Delhasse et al., 2018; Hanna et al., 2012, 2016) and more transient events like atmospheric rivers and cyclonic moisture intrusions (Mattingly et al., 2018; Oltmanns et al., 2019). Surface darkening through upward snowline migration and bare ice exposure (Noël et al., 2019; Ryan et al., 2019), changes in snow and ice optical properties (Gardner & Sharp, 2010), and changes in algae growth (Cook et al., 2020) and other light-absorbing impurities (Tedesco et al., 2016) also increase surface melt. Around the ice sheet periphery, mass loss is also increasing due to enhanced submarine melt from warming ocean waters that can access the glacier fronts via deep fjords, with water mass exchange assisted by wind events (Carroll et al., 2016; Fraser & Inall, 2018; Jackson et al., 2014; Spall et al., 2017). Warming of subpolar North Atlantic waters is associated with enhanced glacier retreat (e.g., Straneo & Heimbach, 2013). Retreat can in turn increase glacier ice loss by further decreasing terminus stability, altering glacier velocities, and producing dynamic thinning (Felikson et al., 2017; Porter, Tinto, et al., 2018).
The notable recent acceleration of change is not only observable across the catchment to continental scale but is also apparent across virtually all climate-connected metrics. Current melt intensity and runoff have increased greatly over the past two decades (e.g., Ahlstrøm et al., 2017; Fettweis et al., 2011; Noël et al., 2019) and are unprecedented for the last 350 years (Trusel et al., 2018) or longer (Graeter et al., 2018). Increased surface melt also helps to drive surface darkening, which has increased significantly over the last three decades (Shimada et al., 2016; Tedesco et al., 2016) and which forms a feedback loop to further enhance melt. The ability of the ice sheet surface to buffer the increased runoff is also decreasing, as firn porosity in which meltwater can refreeze is reduced or cutoff by infilling (van Angelen et al., 2013) and recent firn ice slab development (MacFerrin et al., 2019; Noël et al., 2017; Vandecrux et al., 2019). Where the ice sheet meets the ocean, glacier retreat has increased in recent decades (Carr et al., 2017; Howat & Eddy, 2011), frontal melt rates have increased at least in some regions (Rignot et al., 2016), and changes are evident even in the far north (Hill et al., 2018). Slightly more than half of the mass loss has been attributed to increased runoff in 1992–2018, while the rest of loss was due to increased glacier discharge that doubled during this period (The IMBIE team, 2019). Responding to a combination of surface melting and dynamic thinning, ice surface elevations around the ice sheet margin have also dropped significantly in the 21st century (Csatho et al., 2014; Kuipers Munneke et al., 2015; Porter, Tinto, et al., 2018; Sørensen et al., 2018). Simultaneously, total solid ice discharge increased from ~430 Gt/year during 1986–2000 to ~500 Gt/year since 2005 (Mankoff et al., 2019).
While ice sheet-wide and regional metrics may suggest homogeneous change, the complexities of ice dynamics actually produce highly heterogeneous behavior. This heterogeneity is evident in observations of elevation change (Csatho et al., 2014), surface character (Shimada et al., 2016), discharge (Mankoff et al., 2019), and comparisons of local spatiotemporal behavior (Bartholomaus et al., 2016; Larsen et al., 2016). Changes in ice sheet surface velocities are particularly useful for examining the heterogeneous response to ice sheet-wide mass loss as they provide a direct measurement of the dynamic response of the ice sheet to environmental forcing.
To examine the nature and extent of change across the Greenland Ice Sheet margin from 1985 through 2015, we explore newly available ice sheet surface velocities, published estimates of glacier termini positions, and local elevation change data. We ask the question: What does rapid ice sheet-wide loss of mass look like at the local scale? This research moves beyond previous efforts by exploring local-scale change on an ice sheet-wide scale, combining many previously published terminus advance/retreat data into a synthesized data set, using a newly developed ice sheet surface velocity data set, and exploring the local character of change through combining these data with newly processes surface elevation change data. Exploring the timing and character of change, we inform our understanding of how key ice sheet metrics align with ice sheet-wide trends of ongoing mass loss and visualize the local-scale behavior of an ice sheet as it transitions from a near balanced state to rapid mass loss driven by increased atmospheric and oceanic temperatures and related processes (e.g., Ahlstrøm et al., 2017; Bevis et al., 2019; Cowton et al., 2018; Hofer et al., 2017; Mernild et al., 2011; Porter, Tinto, et al., 2018).
2 Data and Methods
To examine changes in behavior across the full ice sheet, we combine data on ice sheet surface velocities with records of glacier terminus change. For some analyses we focus on the full two-dimensional flow field, while in other cases we use point measurements acquired from a set of 225 individual marine-terminating glaciers (Figures 1d and 2). To highlight the local to regional-scale character of the recent ice sheet reconfiguration, we also introduce data on ice sheet elevation changes.
2.1 Surface Velocity Data
We use ITS_LIVE (Inter-Mission Time Series of Land Ice Velocity and Elevation) velocity data, which is a new National Aeronautics and Space Administration (NASA) MEaSUREs (Making Earth System Data Records for Use in Research) data product (Gardner et al., 2019) that complements the multidecadal ice sheet surface velocity products produced by the Greenland Ice Mapping Project (Joughin, Smith, & Howat, 2017). ITS_LIVE provides ice sheet surface velocity data derived from the full suite of Landsat 4, 5, 7, and 8 satellites, using the auto-RIFT (autonomous Repeat Image Feature Tracking) processing chain (Gardner et al., 2018). Using auto-RIFT, all data are processed in the same grid and using the same method for all images, without resampling, which provides a homogeneous record from which geophysical change can be extracted. The result is a comprehensive record of ice sheet surface velocities since roughly 1985.
Data from Landsat 4 (1982–2001) and Landsat 5 (1984–2013) are sparser in both space and time and have lower radiometric quality than the Landsat 7 (1999 to present) and Landsat 8 (2013 to present) data, all of which affect velocity data quality. Since we are interested in multidecadal changes in ice flow, we reduce the impact of data sparsity by combining data across 5-year periods beginning in 1985 and ending in 2015. When differencing velocities between two pentads, we use the ITS_LIVE reported error to mask out velocity change data based on v > 1.2v_err, where v is the interpentad velocity change and v_err is the root-mean-square error from the combined annual maps for each pentad. We also include all data in areas where v < 50 m/year (regardless of v_err) to support better visualization across areas with little measured change. To compare changes in flow with the background velocity, we use an error-weighted ITS_LIVE composite velocity map with full ice sheet coverage (Figure 1e). Because of the error weighting, the more accurate and numerous Landsat 8 velocity measurements are more heavily weighted in the composite map.
To examine velocity evolution across decades, we difference pentad data for 1985/1990 and 1995/2000, 1995/2000 and 2005/2010, and 2005/2010 and 2010/2015 (Figures 1a–1d and S1–S3 in the supporting information). Differencing the coverage-limited 1985/1990 and 1995/2000 pentads (Figure S1) yields velocity coverage across the east, northeast, and northern ice sheet, as well as the west coast, particularly in the region near Sermeq Kujalleq (also referred to as Jakobshavn Glacier). Data coverage across 1995/2000 to 2005/2010 (Figure S2) is greatly improved, including coverage in the south and more successful measurement of narrow outlet glaciers across the ice sheet. For 2005/2010 to 2010/2015 (Figure S3), data coverage is similar but somewhat improved, particularly capturing more inland observations.
2.2 Glacier Terminus Data
For glacier terminus advance/retreat data, we update the record of glacier terminus position to match our velocity record, mapping the shifting ice sheet edge from the 1980s through the present. As with velocity data, we combine terminus records across 5-year periods from 1985 to 2015. We seek to use as comprehensive a terminus record as possible, combining data from: NASA MEaSUREs annual terminus positions (Joughin, Moon, et al., 2015, 2017), Moon and Joughin (2008), Howat and Eddy (2011, 2020), Bunce et al. (2018), Millan et al. (2018), Murray et al. (2018), and European Space Agency Greenland Ice Sheet data (ESA, 2019). Terminus change was calculated as the change in along-flow distance following the glacier/fjord centerline. To account for heterogeneity in front position change, each terminus position is approximated using a quadratic polynomial. The along-flow frontal position is then defined as the intersection of the quadratic approximation of the terminus and the center flow/fjord line. This approach creates a consistent metric of terminus change that is robust to heterogeneity in frontal shape and gives similar results to the box method (Moon & Joughin, 2008) but without requiring a predefined width approximation from which to calculate changes in glacier area.
2.3 Local Elevation and Subglacial Hydrology Data
Since the ice sheet-wide scale of the velocity and terminus data makes it difficult to visualize and explore local to regional scale implications of ice flow reconfiguration, we include many local-scale examples and a detailed case study from northwest Greenland. For investigating the timing and magnitude of thinning, we use altimetry observations from NASA's ICESat (Ice, Cloud, and land Elevation Satellite, 2003–2009; Zwally et al., 2002, 2014) and ATM (Airborne Topographic Mapper, spring 1994, 1999, 2002, 2005, annually in spring 2009–2015, and also in early fall in 2015; Krabill et al., 2002; Studinger, 2018) and LVIS (Land, Vegetation and Ice Sensor, 2009 and 2013; Blair & Hofton, 2019; Hofton et al., 2008). We extend and densify the altimetry record by incorporating digital elevation models (DEMs) from aerial photogrammetry (23–26 July 1985, Korsgaard et al., 2016a, 2016b), the SPOT 5 stereoscopic survey of Polar Ice: Reference Images and Topographies (SPIRIT) (31 July 2007; Korona et al., 2009), and ArcticDEM strips generated from stereopair image swaths collected by DigitalGlobe's polar-orbiting WorldView-1 and 2 satellites (various dates in 2012–2015; Noh & Howat, 2015; Porter, Morin, et al., 2018). Time series at 19 locations, selected at intersections of ICESat ground tracks and ATM swaths (Figures S2–S8; all elevations are given on the WGS-84 ellipsoid), are generated from altimetry observations and stereo DEMs using the Surface Elevation Reconstruction And Change detection (SERAC) approach (Csatho et al., 2014, 2020; Schenk & Csatho, 2012).
The SPOT DEM is corrected using 18 SERAC time series as control (Schenk et al., 2014). The correction, modeled as a quadratic function, includes a bias of approximately −19 m and a small higher-order correction. The residual error of 1.94 m (root mean square) between the corrected DEM and the altimetry time series is consistent with the expected accuracy (e.g., Korona et al., 2009). In good agreement with previously reported results (Noh & Howat, 2015), the vertical error of most ArcticDEM strips over the ice sheet is estimated to be ~2–3 m (root mean square) based on the comparison of ArcticDEM elevations with time series modeled from the altimetry measurements using piecewise polynomials up to third degree (Figures S5–S7). However, ArcticDEM strips dated 20 and 24 May 2012 had large elevation errors, exceeding 50 m, and thus were excluded from our analysis. To obtain coverage of the entire study site, we mosaic the 2012 and 2015 ArcticDEM strips using ENVI (Exelis Visual Information Solutions, Boulder, Colorado).
We use the 1985, 2007, 2012, and 2015 DEMs to investigate regional changes in surface elevation (between 2007–1985, 2012–2007, and 2015–2012, Figure 4). Potential subglacial hydrologic networks in 1985, 2007, 2012, and 2015 (Figure 5) were reconstructed from the surface DEMs combined separately with the bed topography from BedMachine v3 (Morlighem et al., 2017). After the conversion of the surface DEMs from ellipsoidal to geoidal heights, using the geoid model of BedMachine v3, we calculated the hydraulic potential beneath the ice sheet and generated potential hydrological channel network maps following the methodology outlined by Lewis and Smith (2009).
3 Character of Greenland Ice Sheet Rapid Reconfiguration
3.1 Terminus Position Is a More Consistent Indicator of Environmental Change Than Surface Velocity
Comparing changes in ice motion and glacier termini positions since the mid-1980s, the most consistent trend is widespread terminus retreat, which is apparent across the full ice sheet (Figure 2). Documented in earlier studies (e.g., Carr et al., 2017; Howat & Eddy, 2011; Moon & Joughin, 2008), our combined and updated data set confirms strong glacier retreat particularly since ~2000. For individual glaciers, patterns of retreat are rarely consistent, but the full record of change (Figure 2) demonstrates that most marine-terminating glaciers have retreated substantially over the last several decades (89% of the Figure 1d glaciers have retreated with no seasonal buffer considered). Retreat is most limited in the northern and northeastern regions. This may be a combined result of several factors, including (1) the slow-moving glacier types that dominate these regions (with the exception of the Northeast Greenland Ice Stream), (2) cooler ocean water at depth (Straneo & Cenedese, 2015) that would decrease submarine terminus melt and retreat as compared to other regions, and (3) more limited surface melt that leads to lower rates of terminus thinning (Noël et al., 2016). Lower surface melt also likely leads to lower subglacial runoff at depth, which would reduce the circulation at the ice-ocean boundary, limiting how much submarine melt occurs due to warm ocean water being brought into contact with the glacier ice front (e.g., Motyka et al., 2013). Nevertheless, retreat is pervasive even in these areas of more limited change. And while the terminus position of some marine-terminating glaciers has remained roughly stable, substantial sustained advance is virtually absent across the entirety of the ice sheet.
While glacier terminus observations reveal relatively consistent and widespread retreat, changes in glacier motion encompass a much wider range of responses, including both substantial speedup and slowing (Figure 2). With the shift away from mass equilibrium between 1985/1990 and 1995/2000, the most intense speedup is apparent at Sermeq Kujalleq on the west coast (Figure 1b) and Kangerdlugssuaq Glacier on the east coast (Figure 1a). More muted speedup is observed at most sampled west and northwest glaciers, while several northeastern glaciers slowed somewhat, and many glaciers across the east saw little change. From 1995/00 to 2005/10 (Figure S2), trends of both speedup and slowing are more intense than in the earlier period. Rather than consistent regional trends across marine-terminating glaciers, heterogeneous behavior emerges across the ice sheet. Locally and regionally heterogeneous velocity trends remain the norm for 2005/2010 to 2010/2015 (Figure S3), and the multidecadal trends (Figure 2) demonstrate that long-term ice flow behavior covers a large range from slowing to speedup.
The diversity of glacier velocity responses from slowing through speedup should not be surprising. The full range of slopes and geometry of bedrock and sediments that underly these glaciers can in itself facilitate both speedup and slowdown as a response to frontal retreat. In addition, the expected response can vary over time as the terminus retreats and advances, becoming influenced by different subglacial topographic features. These topographic controls can create short-term and long-term periods of speedup and slowing that can be unrelated to instantaneous terminus forcing (e.g., Carr et al., 2013; Catania et al., 2018; Enderlin et al., 2013). As a result, long-term trends in glacier retreat (which can still occur episodically) provide a more robust indicator of ice sheet-wide response to contemporary changes in climate than do changes in surface velocity or solid ice discharge. Remotely sensed observations of ice sheet surface elevation change, now supported via CryoSat-2 and ICESat-2 and ongoing DEM development from satellite data, are also powerful indicators for ice sheet mass loss. We demonstrate the utility of combining elevation data with terminus position and velocity observations in section 3.2.1 but have not examined the strength of elevation change as an indicator on an ice sheet-wide scale.
3.2 Local Character of Rapid Reconfiguration
Previous studies have primarily focused on the broader surface velocity trends (e.g., Moon et al., 2012). This limitation was in part because resolving flow changes along the glacier edge or confidently measuring areas of slow or slowing velocity has been difficult due to limitations that included short remote sensing time series, measurement interference from fjord topography or shadows, and limits due to instrument design. Multidecadal records and advances in remote sensing technology and data analysis methods now allow us to examine the more complex patterns of change that are reshaping the Greenland Ice Sheet periphery. These changes include narrowing of fast-flow zones, ice rerouting, and likely outlet abandonment or piracy, in which flow of new ice to a particular glacier slows substantially, effectively leaving that outlet glacier stranded and changing in place due to surface mass balance. Because glacier dynamics are closely tied to local variables like topography and climate, however, these indicators of change are not present on all glaciers, and some glaciers show only one or two of these indicators. To elucidate the character of these indicators, we describe the changes in more detail for several local areas.
3.2.1 Northwest Case Study
To illustrate the rapid reconfiguration underway along the Greenland coast, we zoom into northwestern Greenland, focusing on an area around Kjer and Hayes Glaciers (Figures 3-5 and S4–S10). Many indicators of ice sheet reconfiguration are apparent in this region, including intensified ice flow with some ice stream narrowing, potential rerouting of ice, and likely outlet abandonment.
All glaciers in this area have retreated, with accelerated rates toward present. While 1985/1990 to 1995/2000 velocity changes are more subtle (Figure 3a), they set the stage for the multidecadal pattern of change, with ice speeding up for the primary outlets of Kjer Glacier (near elevation time series point P1) and Hayes Glacier (near P10) and slowing for the northernmost outlet (near P2) and central outlet (near P7) (point identifiers are available in Figure 4a). This pattern continued through 2005/2010 (Figure 3b) with more pronounced slowing for most of the central P7 outlet, although a small southern portion of it sped up, likely in response to 2000–2005 retreat that brought this portion of the glacier front to the 2015 position. The main outlet of Hayes (near P10) sped up. From 2005/2010 to 2010/2015 (Figure 3f), slowing continued near P7, including the small area of previous speedup, and speedup intensified for the Kjer and Hayes outlets. Changes in this region mirror trends across the ice sheet: widespread retreat combined with a locally heterogenous velocity response (see Figure 4d for changes from 1985/1990 to 2010/2015).
Pulling in elevation data provides area DEMs (Figures 4a–4c) and point measurement time series (Figures S5–S10) that we divided into three regions according to the ice flow pattern, Kjer Glacier (P1–P6, P13, P15, and P18), central outlet (P7–P9), and Hayes Glacier (P10–P12, P14, P16, P17, and P19). Thinning dominates across the full region over the entire study period, though there are marked local differences in thinning rates, including transient periods of local thickening or little change.
Prior to 2007, most thinning occurred in the northern half of the study site within the drainage basin of Kjer Glacier (Figure 4a). Average thinning rates determined from the 19 time series featured in the study were −1.7 m/year in the Kjer Glacier, −0.9 m/year in the central, and −0.6 m/year in the Hayes Glacier basins. For the Kjer Glacier and its tributaries the relationship between the ice sheet elevation and elevation change can be modeled with a linear function resulting in a thinning rate of −3.5 m/year around sea level and a lapse rate of −0.36 m/year for every 100-m elevation change (Figure S8). Thinning was small and spatially uniform on Hayes Glacier.
Between 2007 and 2012, thinning accelerated in the entire region (Figures 4b and S9) with average thinning rates increasing to −2.8, −1.7, and −4.7 m/year over Kjer, central, and Hayes Glaciers, respectively. Noteworthy is the onset of very rapid thinning in the lower region of the Hayes Glacier (P10 and P19 below the ice fall and P12 over a small fast flowing region above the ice fall). At the lower elevations (P9 and P10) thinning rates increased more than 10 times between the 1985–2007 and 2007–2012 periods. Thinning rates slightly decreased near the terminus of Kjer Glacier (P2 and P3), suggesting that the glacier started to adjust to a new equilibrium (Figure S5). The long-term SERAC time series, which include repeat ATM altimetry in the 1990s at P1 and P18, also indicate a slight decrease in thinning rates in the early 2000s (Figure S5). Thinning rates change linearly with elevation on both the Kjer and Hayes Glaciers with a thinning rate of −8.7 m/year near sea level and lapse rate of −1.1 m/year for every 100-m elevation change (Figure S9). Thinning rates remained low in the central outlet region.
Rapid thinning continued in 2012 and 2015 with a spatial pattern similar to 2007–2012 (Figures 4c and S10). Thinning rates doubled at lower elevation both on Kjer and on Hayes Glaciers, reaching −16 m/year at sea level, while thinning ceased at the higher-elevation regions surrounding the outlet glaciers. Thinning of Kjer Glacier accelerated and extended to its tributaries (Figure 4c). At the same time, most thinning occurred at lower elevations, near and below the ice fall on Hayes Glacier.
Most time series in the study (Figures S5–S7) show an initial steady decrease of ice sheet elevation between 1985 and 2003–2005, followed by the onset of accelerating thinning on both Kjer and Hayes Glaciers around 2010. The increase of thinning was slow and gradual in the central region (Figure S6). The sudden increase of thinning rates at P14 and P16 on Hayes Glacier in 2010 might be indicative of a redirection of ice flow, from the west-northwest to east-southeast trough located under the central region toward the Hayes Glacier ice fall. The most rapid thinning is observed at P19, which thinned ~90 m in 5 years. Finally, P1 was just upstream of the terminus of Kjer Glacier in 1985. By March 2012 the glacier retreated beyond this location, and thus, its time series depicts the sea level with occasional icebergs after 2012.
Thinning at Kjer Glacier is roughly coincident with terminus retreat (Figures 4a–4c), fulfilling a classic dynamic feedback between retreat and speedup particularly for glaciers with deep troughs and retrograde slope (see Figure 5d for bed topography). As retreat has begun to cutoff the northernmost outlet, ice flow to that outlet has slowed, and the outlet has been effectively abandoned.
The pattern of change in the southern half of the area is more complex. Despite a long, deep southeast to northwest trough (Figure 5d) running past Hayes Glacier, flow here is routed into Hayes rather than the central outlet.
Exploring potential subglacial meltwater pathways using 1985, 2007, and 2015 DEMs and BedMachine v3 ice thickness (Figures 5a–5c) suggests subtle but notable change as a result of thinning. In 1985, the primary channel underneath Kjer may have discharged on the northern edge of the glacier (Figure 5a). This outlet may have been abandoned by 2007 (Figure 5b) and further compromised by terminus retreat to 2015 (Figure 5c). In the southern region, the subglacial network underneath the Hayes outlet has shown little to no connection to the channel system aligned with the deep southeast-northwest trough at any point in our record. Despite more minor retreat in the central outlet, the discharge location for this central hydrologic network may have shifted from 1985 to 2007 and again by 2015. Some channel paths in both north and south regions also became more pronounced by 2015.
The potential implications of channel rerouting deserve expanded and in-depth research; they include changing access to subglacial sediment, alterations in subglacial erosion, or basal water pressure-induced sliding and shifting locations of meltwater discharge that can influence calving style and rate (e.g., Cowton et al., 2019; Fried et al., 2018) and local ocean properties (e.g., Kanna et al., 2018; Seifert et al., 2019).
3.2.2 Examples From Across the Ice Sheet
Narrowing fast-flow zones are apparent in locations across the ice sheet. Rink Glacier along the western coast provides an example (Figure 6). Between the late 1980s and the late 1990s, most of the fast-flowing area of Rink sped up (Figure 6a). Into the late 2000s, speedup near the glacier midline continued, but ice along the shear margins slowed (Figure 6b). Ice along the midline did slow into 2010/2015 but did not return to 1985/1990 speeds (Figure 6c). As a result, the full pattern of flow changes from 1985/1990 up to 2010/2015 shows pronounced slowing along the glacier edges and speedup along the center flowline, narrowing the zone of fast glacier flow (Figure 6d). These dynamic changes have occurred even as Rink's terminus has remained relatively stable due to a combination of fjord geometry (Catania et al., 2018) and a discharge flux that is much higher than the submarine melt rate (Rignot et al., 2016), while dynamic thinning and mass loss have been spatially variable (both positive and negative) across the catchment (Felikson et al., 2017). It is valuable to note that instrument sensor errors are more likely to produce errant signals of shear margin speedup, the opposite of our observations, increasing our confidence that these satellite observations are resolving real ice flow changes (Dehecq et al., 2018). Glacier margin slowing combined with speedup along the primary flow channel is evident for other glaciers both with longer fjords like Rink, such as two north-central outlets of Upernavik Isstrøm (Figure S11), and also evident for glaciers that do not flow through such well-confined fjords, such as in the Ikertivaq region of southeastern Greenland (Figure S12).
Patterns of speedup and slowing that suggest flow rerouting and outlet abandonment are also common across the ice sheet margin. Limits in confidently calculating small changes in ice flow direction from remotely sensed data, however, make assessment of these indicators more speculative. Bed topography that favors flow rerouting under conditions of terminus retreat and glacier thinning raise confidence that specific areas will continue to slow. For example, Sermeq Silarleq on the western coast is speeding up, while Perlerfiup Sermia, which diverges north from Sermeq Silarleq over a topographic high, is slowing down and retreating (Figure S13). The topographic setting suggests little opportunity for Perlerfiup Sermia to regain earlier velocities or terminus position. In contrast, the topographic setting in other locations suggests that future speedup is possible, such as at the south-central outlet of Upernavik Isstrøm (Figure S11) (see also Larsen et al., 2016).
4 Implications and Conclusions
Using several multidecadal observational records, we elucidate the local-scale changes that act as indicators of rapid Greenland Ice Sheet margin reconfiguration as a result of substantial and ongoing mass loss. Retreat is widespread, while ice flow responds across a broad range of speedup and slowing due to local topographic features. Combined records of glacier terminus position and ice velocities emphasize that slowing motion is not necessarily an indicator of stability. The glaciological community should be cautious in suggesting speedup as a consistent response to retreat (even if it is a dominant response) or as an indicator of a Greenland climate response. Instead, pan-Greenland reconfiguration of ice flow, which includes narrowing fast flow zones, ice flow rerouting, and potential outlet abandonment, acts as an ice sheet-wide indicator of rapid and widespread change. Ice sheet reconfiguration is expected to influence iceberg discharge, water (and sediment) routing, glacier basin boundary migration, and both short and long-term trends in flow regime.
The pervasive nature of the observed reconfiguration and projections of sustained increases in air temperature (IPCC, 2013) and subsurface ocean warming (Yin et al., 2011) suggest that overall rapid reconfiguration is likely to continue. We find no indication that recent changes are temporary. Rather, we expect that these local-scale changes, which are playing out across the ice sheet, demonstrate the character of any rapid ice sheet retreat and may be a useful analogue for understanding both past and future ice sheet changes.
For informing understanding of past deglaciation, the types of patterns we identify here—and the extensive underlying data sets that we use and highlight—may be helpful for assessing exposure rates as an ice sheet edge retreats and expected dynamic behavior (e.g., to inform understanding of Greenland retreat from the Last Glacial Maximum, Batchelor et al., 2019). To improve the utility of using contemporary observations as an analogue for previous deglaciation, research should consider whether the rapid pace of contemporary change fundamentally alters any primary mechanisms of ice sheet reconfiguration. For example, would slower warming facilitate a different elevational profile evolution because glaciers would have longer response periods for slower increases in air or ocean temperatures?
Our observations and the data sets highlighted within this paper may also be useful in projecting future changes in the ice sheet margin and ice dynamics. While ice sheet models are advancing in capability, they are only recently able to capture some of the complex flow observed with satellites in outlet glacier systems near the coast (Aschwanden et al., 2019; Morlighem et al., 2019), and many local-scale flow features are still not captured in models (Mottram et al., 2019). Potentially important processes such as the influence of light-absorbing impurities (Tedesco et al., 2016) and biota (Cook et al., 2020), and dynamics that we are still aspiring to fully understand, like the relationship between ice sheet speeds and subglacial hydrology, and evolving basal friction (Minchew et al., 2019; Stearns & van der Veen, 2018, 2019), are also still missing or in development for most ice sheet models. This limits the utility of full ice sheet models for predicting local changes on decadal timescales, and it may be some time before ice sheet models can provide this type of future projection with high confidence. In the interim, multidecadal observations like those explored here, combined with topographic and climate data, may provide a path forward for projecting the next several decades of Greenland Ice Sheet evolution.
Acknowledgments
T. M. was supported by NASA GoLIVE grant NNX16AJ88G and USGS Landsat Science Team grant 140G0118C0005. A. G. was supported by the NASA Cryospheric Sciences Program and the NASA MEaSUREs program through the ITS_LIVE project. M. F. was supported by the NASA Cryospheric Sciences Program through NASA GoLIVE grant NNX16AJ88G and the NASA MEaSUREs program through the ITS_LIVE project. B. C. and I. P. were supported by NASA's Operation IceBridge Science Team grant NNX17AI65G. Acknowledgement to NASA MEaSUREs ITS_LIVE project for publication funding. We acknowledge Anna Covey for terminus data processing, Ian Howat for terminus data contributions, Tony Schenk for calculation of SERAC time series and SPOT DEM correction, and Leigh Stearns for helpful discussion.
Open Research
Data Availability Statement
There are no restrictions to access for any data within this manuscript. Data sources are cited in the main text, with the access information further summarized here. ITS_LIVE ice sheet surface velocities available at the website (https://its-live.jpl.nasa.gov/#data). Terminus data sourced from Bunce et al. (2018), European Space Agency (2019), Howat and Eddy (2011, 2020), Joughin, Moon, et al. (2015, 2017), Moon and Joughin (2008), Millan et al. (2018), and Murray et al. (2018). SERAC surface elevation time series are available at the website (https://zenodo.org/record/3665445#.XkQ-1mhKi70), including additional metadata on the data source files. DEMs were used from the websites (https://doi.org/10.7289/v56q1v72 (NCEI Accession 0145405) and https://www.pgc.umn.edu/data/arcticdem/uncorrected, and https://theia.cnes.fr/atdistrib/rocket/#/search?collection=Spirit), modified by a nonlinear transformation determined from the altimetry time series.
