Introduction

The Greenland ice sheet has been one of the largest contributors to sea-level rise during the past two decades (IPCC, 2013). Large-scale changes and increased mass loss have been observed in many regions of the Greenland ice sheet. The rates of mass loss have varied regionally with the most negative trends coinciding with regions where ice is discharged through marine-terminating outlet glaciers, which have thinned, sped up, and retreated (Pritchard et al., 2009; Rignot and Kanagaratnam, 2006; van den Broeke et al., 2009; Joughin et al., 2010; Moon et al., 2012). Iceberg calving from marine-terminating glaciers accounts for between 36% and 58% of Greenland's overall mass loss in the period 2000–2009 (Enderlin et al., 2014), and calving is known to dominate mass loss in regions of the ice sheet such as the southeast (SE) and northwest (NW) (e.g., Pritchard et al., 2009; van den Broeke et al., 2009). Changes in the calving-front position of marine-terminating glaciers are a good indicator for dynamic mass loss, as typically speedup-induced thinning is coupled to frontal retreat (Howat et al., 2005, 2008a; Joughin et al., 2008).

Studies of calving-front positions (or studies that include such data) have used a small number of temporal snapshots of ice-front positions (Moon and Joughin, 2008; Howat and Eddy, 2011), have been regionally specific (Howat et al., 2008a; Joughin et al., 2008; McFadden et al., 2011; Murray et al., 2010; Seale et al., 2011; Walsh et al., 2012), or have only sampled a small number of glaciers (Box and Decker, 2011). In what is one of the most extensive published studies, McFadden et al. (2011) considered 59 glaciers in west Greenland, and this study showed no apparent synchronicity in glacier behavior in this region of the ice sheet. In contrast in the SE, synchronicity of retreat and subsequent restablization/advance of the fronts of tidewater glaciers during the 2000s has been reported (Rignot and Kanagaratnam, 2006; Howat et al., 2008a; Murray et al., 2010; Walsh et al., 2012) and attributed to regional forcing such as air or ocean temperature changes (e.g., Luckman et al., 2006; Moon and Joughin, 2008). Differences in the behavior of the calving fronts of glaciers north and south of 69°N in the SE has been suggested to be controlled by the patterns of ocean circulation (Seale et al., 2011; Walsh et al., 2012).

However, to date no ice-sheet-wide study has been published that quantifies calving losses with both high spatial and temporal coverage. In this paper, we present the first complete time series of terminus fluctuations for all Greenland tidewater glaciers wider than 1 km that are visible in Landsat data in order to resolve both the seasonal and inter-annual changes during the period 2000–2010. We also investigate various environmental conditions in each region thought to be plausible as explanatory controls on glacier calving and frontal position: namely sea surface temperature (SST), melt days, and sea ice coverage.

Methods

We compiled a record of terminus positions for all tidewater glaciers with width larger than 1 km visible on Landsat-7 data around the Greenland ice sheet during the 11-year period 2000–2010 (Fig. 1), thus capturing 199 tidewater outlets in a time period where extensive changes have been reported (Howat and Eddy, 2011). In common with all studies of this nature, a cut-off size below which fronts were not considered was needed to enable digitization of the fronts within a practical time scale. The orbital limit of Landsat-7 data is 81°N (NASA, 2014), although in some places data up to 81.7°N are obtainable. This data limit means that some regions of northern Greenland are sparsely or not covered by the data, but the majority of Greenland's coast has good coverage. Ice caps in contact with the Greenland ice sheet, such as the Geikie Plateau (GEP in Fig. 1) are taken as part of the main ice sheet. We omit several outlets along the GEP known to be surge-type glaciers because this region is a surge cluster (Jiskoot et al., 2003). The frontal positions of surge-type glaciers respond to an internally driven cycle rather than climatic changes (Clarke, 1987) and a region with a high percentage of surge-type glaciers could affect the results. Two further surge-type glaciers, one in the NW and one in the northeast (NE) (Harald Moltke Bræ [76.6°N, 67.85°W] and Storstrømmen [76.7°N, 22.45°W] [Rignot and Kanagaratnam, 2006]) were included. No new surge-type glaciers were identified during our analysis, however, the data set likely also contains a small number of unrecognized surge-type glaciers. Furthermore, the record of the Northeast Greenland Ice Stream (NEIS in Fig. 1) comprising the outlets Nioghalvfjerdsbræ and Zachariæ Isstrøm was not included in our analysis due to ambiguities in identifying the actual calving cliff locations in imagery.

FIGURE 1.

Location of the glaciers included in the data set. The circle size indicates the number of measurements between 2000 and 2010. Note that circle symbols are plotted with the largest symbols lowest and smallest symbols on top, meaning that there are no fully obscured symbols on the figure. Dark green indicates glaciers with more than 1500 m retreat (which form those glaciers in the red class in Fig. 6). Numbers are the total number of glaciers in each region of the ice sheet, GEP is the Geikie Plateau region, NEIS is North-East Greenland Ice Stream.

We used Landsat thematic mapper (TM) and enhanced thematic mapper (ETM+) images downloaded from the National Aeronautics and Space Administration (NASA). These images were orthorectified by the method presented by Tucker et al. (2004) and geolocation errors were minimized by using only data of the same Worldwide Reference System (WRS) path/row for each glacier. Furthermore, we visually inspected the alignment of several images for each glacier along the steep fjord-walls, and found no image with irregularities larger than one pixel, translating on the ground to <30 m for Landsat-5 TM, and <15 m for Landsat-7 ETM+ (band 8). Glacier-front positions were then manually digitized on-screen from the Landsat scenes. Since 2003, a hardware fault in the Scan-line Corrector of the Landsat-7 satellite means that there is an ∼22%–25% data loss in the form of parallel stripes varying in width across each image (Wijedasa et al., 2012), however, it was still possible to map calving front positions from these images: front positions were simply digitized either side of the missing data.

There are a number of methods by which glacier frontal change can be calculated (e.g., Walsh et al., 2012; Lea et al., 2014): all have limitations, some of which are detailed by Lea et al. (2014). The commonest method is the center-line method; however, more recently the more complex and time-consuming box method has been increasingly used (e.g., McFadden et al., 2011; Moon and Joughin, 2008; Howat and Eddy, 2011), because of its ability to account for asymmetric changes in front geometry. However, the box method is problematic unless the box width is equal to the glacier width (Lea et al., 2014), and so is inappropriate for data where segments of fronts are regularly missing or where the glacier front changes width: both occur in our data. We chose to use a modified center flow-line method where the flow-line follows the fjord orientation. This center flow-line method of measuring frontal positions was shown by Walsh et al. (2012) to be equivalent in accuracy (to ±100 m, which is the measurement error) to the box method. In this study, center flow-lines were manually drawn and where necessary segments followed changes in fjord orientation. We considered the width of the respective outlet, its surface velocity field (Rignot and Kanagaratnam, 2006) close to the calving front, as well as choosing a more-or-less perpendicular orientation compared to the digitized calving fronts. Using this modified center flow-line method, the distance between front positions was measured along the center flow-line of each outlet relative to the position of furthest retreat. Measurement accuracies on the flowlines are estimated to be better than ±76 m (Bevan et al., 2012), and each center-line location was chosen to avoid any gaps due to the Scan-line failure.

To validate this choice of method we randomly chose 10 glaciers and compared Lea et al.'s (2014) extrapolated centerline method with our method for all measured fronts at these glaciers. The extrapolated center-line method uses inverse distance weighting to extrapolate positions from the glacier center line across the fjord width and allows for both changing fjord width (unlike the box method) and frontal geometry (Lea et al., 2014). The median difference between frontal changes using the absolute value of retreat or advance using our method and the extrapolated center-line method was 77 m and the interquartile range was 107 m across 357 measured frontal positions. Furthermore, neither method was demonstrably better than the other: our center-line method does not deal with changes in acrossfront geometry (such as a glacier front changing from concave down-glacier to convex down-glacier), but the extrapolated center-line method is problematic when fronts are sometimes partially obscured by cloud or the Scan-line hardware fault, as is the case in our data.

In order to ensure measured frontal changes are not in fact simply seasonal variations, we aimed for a minimum of three measurements per year (covering spring, summer, and autumn whenever possible, although limited imagery prevents a regular seasonal survey), subject to satellite data coverage (Fig. 1). We discuss regional seasonal variations in the frontal position of the glaciers and their timing; however, there are limitations on these results and they should be taken as minimum values of seasonal variation as it is unlikely that our sampling records either the maximum or the minimum position of individual glaciers. We grouped the outlet glaciers into five sectors around the ice sheet (Fig. 1) based on previous studies (e.g., Luthcke et al., 2006; Wouters et al., 2008; van den Broeke et al., 2009; Sasgen et al., 2012). We incorporated Kangerdlugssuaq Gletscher (68.6°N, 32.48°W) into the SE sector, as it is known to have behaved in concert with other SE Greenland glaciers (Luckman et al., 2006; Howat et al., 2008a; Murray et al., 2010). Figure 1 shows the number of glaciers in each region and summarizes the overall number of measurements we made at each glacier. On average there are two measurements per year for glaciers in the north (N) sector, four in the NE, three in the SE, two in the southwest (SW), and four in the NW. There are only two glaciers in the database with measurements completely missing in any particular year (i.e., zero measurements in a year): for both glaciers this occurs for just a single year. Nevertheless, because we present averaged positions, there are 45 occurrences within the database without a recorded annual frontal position because fewer than three measurements could be made in that year, and one glacier (Steensby Gletscher, 81.6°N, 54.64°W) has no average position for the five-year period 2000–2004 (Table 1). The orbital limit means many more of the glaciers with missing averaged positions are in the northern sector than are further south, which together with the small number of glaciers measured in the N region (Fig. 1) means that results in the N can be affected by the variable data coverage. In total, the inventory comprises 6688 ter minus positions with satellite coverage varying significantly for the different areas around the ice sheet (Fig. 1). Values when stated for a particular year are calendar years (1 January to 31 December). In this paper, we present regional summaries of the data collected: information for all individual glaciers is presented in Appendix Table A1, and the locations and fronts themselves are presented in a Google Earth Keyhole Markup (.kmz) file provided as  Appendix supplementary file A1 (427-447 pp AAAR-47-3-02 supplementary file A1.kmz), which is available online as an open access file.

TABLE 1.

Glaciers that advance or are stable relative to their first recorded position during the period 2000–2010. Steensby Gletscher (81.6°N, 54.63°W) is also recorded in the database as advancing 186 m during the period 2005–2010, but there are no frontal positions for this glacier within the database prior to 2005.

FIGURE 2.

Histograms showing number of occurrences of annual frontal change of different magnitude in: (a) northwest; (b) north; (c) northeast; (d) southwest; and (e) southeast regions. Figure 1 shows the regions and locations of glaciers measured, along with the number of measurements. (f) Histogram of annual frontal change for all glaciers measured. Note that in all cases the horizontal axis is logarithmic and that vertical axes vary between regions. n is the number of glaciers in each region. Advances in blue, retreats in green.

Because the frontal changes measured are not normally distributed (small changes are more common and the distributions are positively skewed; Fig. 2), both the mean and median values of retreat for the glaciers in each of the different sectors were computed: in order to achieve this, a daily frontal record was interpolated from the measured data points for each glacier and used to calculate the mean or median position across each sector. For one figure, glaciers within each sector were also separated by the overall magnitude of their frontal change. In common with analysis in Murray et al. (2010), those glaciers with overall retreat exceeding 1500 m over the time period were considered “large” retreats, and those with retreats less than 1500 m were considered “small” retreats (Fig. 1). All overall advances (Table 1) were less than 1500 m and therefore were considered “small.”

We also examined possible environmental controls on glacier frontal position, specifically sea surface temperature, melt days, and sea-ice extent.

A number of studies have shown or implied that the delivery of heat by ocean water to the fronts of calving glaciers in Greenland is an important control on frontal retreat (e.g., Holland et al., 2008; Murray et al., 2010; Straneo et al., 2010; Christoffersen et al., 2011), whether by weakening the ice mélange (Amundson et al., 2010) or undercutting the ice front directly (Motyka et al., 2003). Satellite-measured sea surface temperature (SST) has been used as a proxy for this heat, albeit that it reflects the temperature of the very surface waters. In this paper, we present Moderate Resolution Imaging Spectroradiometer (MODIS)-derived SSTs (Brown and Minnett, 1999) for comparison to glacier behavior. SST in the SE has been used to explain glacial retreat (Howat et al., 2008a; Hanna et al., 2009; Murray et al., 2010). An analysis by Sutherland et al. (2013) has subsequently shown SST is significantly correlated with water temperatures in the upper 50–250 m; however, there is no significant correlation at greater depth. SST may therefore only be a suitable proxy for the integrity of the ice mélange rather than the influx of deeper warm water. In the SE only, we also use temperatures measured at depth by using the westernmost station (64.2°N, 27.58°W) of the quarterly measured Faxaflói-line, an Icelandic long-term hydrographic standard section ( http://www.hafro.is/∼argos/snid/snid.php?stod=FX9&dypi=200). This station captures the core of the Irminger Current (IC) prior to its recirculation southward and before it interacts with the waters on the East Greenland shelf. We use mean values of temperature for a depth range of 200– 500 m (H. Valdimarsson, personal communication) to capture the full depth of the IC while minimizing any seasonal influence.

We also used the Greenland Daily Surface Melt 25 km EASE-Grid 2.0 Climate Data Record from NSIDC to compute melt day anomalies for each catchment (Mote, 2012), essentially summing the number of days on which melt occurs in each catchment as a proxy for melt intensity. Finally, the coverage of sea ice in the coastal region was taken from the Sea Ice Index (Fetterer et al., 2002).

Results

In general, Greenland's tidewater glaciers experienced large and widespread retreat during the period 2000–2010 (Fig. 2): retreat is dominant around the entire ice sheet margin. In total, the 199 outlets measured retreated more than 267 km over the 11-year period (Table 2; Fig. 3). The largest annual retreat of any glacier (14.9 km) occurred at Hagen Bræ in N Greenland (81.5°N, 28.72°W). Annual rates show strong variation, with maximum total retreat in 2009 (-44 km), 2004 (-42 km), and 2005 (-41 km), whereas there was an overall advance in 2006 (+4 km) (Figs. 4 and 5). However, only 11 of 199 outlets show minor advance over the whole 11-year period (Fig. 2; Table 1). There is no obvious regional clustering of the glaciers that advance and they occur across four of our five sectors.

TABLE 2

Summary results for period 2000–2010 by sector. Columns 4–7 are normalized by the number of glaciers in the sector.

The mean retreat rate per glacier per year was ∼355 m a^-1 in the N, ∼145 m a^-1 in the SE, ∼130 m a^-1 in the SW, ∼115 m a^-1 in the NW, and ∼45 m a^-1 in the NE. However, because the number of glaciers in each sector differs, the NW was the main contributor to the overall retreat, followed by the SE, and SW (Fig. 4, part a). There is considerable inter-annual variability in both the overall retreat rate and the regional pattern of retreat (Table 3; Figs. 4 and 5, part b). Overall retreat rates in Greenland were highest during 2009 (Fig. 5, part b). During 2009, no region showed overall advance and the northern glaciers retreated strongly (due primarily to the retreat of Hagen Brae). However, only the northern glaciers were retreating at their fastest rate during this year. Overall retreat rates were almost as high in 2004 and 2005 despite the northern glaciers advancing slightly in those years (Fig. 5, part b). Glaciers in Greenland showed overall advance during 2006 (Figs. 4 and 5, part b), driven by low retreat in all sectors with a slight advance in the NE and strong advance in the SE. Other years with overall low retreat rates were 2001, 2008, and 2007: glaciers in the SE continued advancing into 2007 (Fig. 4).

Regionally, glaciers in the NW showed overall retreat every year in the record, whereas all of the other regions had at least one year with advance (Fig. 5, part b). The glaciers in the NW retreated most strongly of all the regions in 6 out of the 11 years in the record, with 4 of these years occurring since 2006. The NW also contained the largest number of glaciers retreating more than 1500 m during 2000–2010. Glaciers in the SE retreated more than those in the other regions in 2002, 2005, and 2009, with the SW dominating in 2003 (Fig. 5, part a). Glaciers in the SE showed the greatest variability in frontal position: both the greatest overall retreat (during 2005) and greatest overall advance (during 2006) are shown by the glaciers in this region. The region contained the second largest number of glaciers retreating more than 1500 m over the period.

Figure 6 shows both the inter- and intra-annual variability in glacier frontal position in each region broken down by size of retreat. Seasonal variations in frontal position are evident in all regions except the north. In those regions with seasonal variation, the pattern revealed is slow advance of glaciers during the winter and more rapid and shorter-lived retreat during the summer period. This pattern is revealed in both the mean and the median behavior for regions (Fig. 6). Seasonal changes are larger for the glaciers with larger overall retreats.

Glaciers in the NE with small overall retreats did so at an almost constant rate, punctuated by seasonal fluctuations (Fig. 6). Those outlets with large retreats showed higher retreat rates during 2002–2005 with subsequent stabilization (Fig. 6). A similar pattern to the NE (rapid retreat to a minimum in 2005, followed by stabilization or readvance) is apparent in the SE across both size classes. Smaller retreats in the SW also follow a similar pattern; however, here the glaciers are most retreated in 2004. Larger retreats in the SW are dominated by the behavior of Jakobshavn Isbræ, which retreated rapidly in the early part of the record, causing the mean and median retreat rates to differ substantially. Seasonal variations are increased after 2005, and much of the advance in the latter part of the period results from the lack of summer retreat during 2008. In contrast, in the NW the outlets with larger retreats show a rather constant mean and median retreat superimposed by seasonal fluctuations, while the outlets with smaller retreats reduced their retreat after 2004. Both size classes increased the magnitude of their seasonal fluctuations 2005 onward (Appendix Table A2).

Figures 7, b–e, summarize the overall potential environmental controls in comparison to the glacier frontal record (Fig. 7, part a), although none of these appears to correlate strongly with overall frontal position. Pearson product-moment correlation shows no relationships between these environmental controls and frontal position that are significant at the 95% significance level. If just the SE glaciers are considered, there is a significant correlation between the SE glaciers frontal position and the temperature of the IC lagged by one year (Pearson correlation—0.70 significant at 95% confidence level).

Figure 7, part e, shows melt day anomalies for each region. The highest melt day anomaly occurs during 2010 in the SW, although all other areas except the north were also affected (Fig. 7, part e, Tedesco et al., 2011). Other high melt years were 2002 and 2008 (N and NE), 2005 (N, NE, NW), and 2007 (SW, SE, NW). Low melt years were 2001 (NE and SW) and 2006 (all areas).

In the SE, SST was high in 2003 and low in 2002 and 2006 (Fig. 7, part d). SST in SW Greenland follows similar trends in most years: both the SE and SW are affected by warm water from the Irminger Current flowing southward along the SE Greenland shelf edge and subsequently northward along SW Greenland (e.g., Straneo and Heimbach, 2013). Cooler waters in the SW during 2009 provide an exception as the SE is relatively warm in that year. SST in the NE is warmest during 2002 and 2008.

Figure 7, part f, shows that inter-regional differences in sea ice coverage are much larger than temporal intra-regional differences. There are sea ice lows in the SE during 2003 and 2005 that correlate to higher SST, but coverage in the SW does not follow a similar trend. Coverage in the SW is low 2004–2006 and peaks in 2008.

Discussion

Overall, Greenland tidewater glaciers experienced large and widespread retreat during the observation period 2000–2010 with strong regional and temporal variations around the ice sheet. Studies of ice sheet mass balance that use a surface mass balance minus discharge method assume a constant grounding line or front position, which given the ubiquitous retreat around Greenland is clearly a potential source of error. Our data could be used to assess the impact of such an assumption.

FIGURE 3.

Overall change in frontal position of sampled glaciers for the period 2000–2010. Note that symbols are plotted with the largest symbols lowest and smallest symbols on top, meaning that there are no fully obscured symbols in the figure. The advances in blue are plotted on top of retreats (in red) where they fall in the same size category.

Dividing the ice sheet into five sectors based on glaciological considerations and published GRACE measurements (Sasgen et al., 2012), a clear pattern with high retreat rates during 2002–2005 and subsequent stabilization becomes clear for the NE, SE, and SW sectors, while outlets in the NW show a relatively constant retreat over the entire period (Fig. 6). The pattern of temporally consistent retreat in the NE and around the southern region of the ice sheet suggests related controls on glacier behavior, possibly originating from warming North Atlantic ocean waters (e.g., Holland et al., 2008; Straneo and Heimbach, 2013). The correlation between the front positions in the SE and the IC temperature reinforces this suggestion.

TABLE 3

Median glacier change in meters for each year by sector. Yearly values in the north where the number of glaciers is small and there are few images available are based on small samples and are therefore not presented.

There is also a clear seasonal signal in all areas except the north showing fast retreat in summer and slow advance during winter (Fig. 6). Some caution should be made in interpreting this result, as the lack of daylight means that Lands at data are not available during winter months in northerly locations. However, a similar pattern of summer retreat and winter advance was found for five of the largest glaciers in Greenland (Schild and Hamilton, 2013). The seasonal signal suggests that the glacier fronts are stabilized by the presence of sea ice or lack of surface melting in winter. Summer retreat is likely initiated by melting causing some combination of sea ice and mélange breakup (Amundson et al., 2010), faster glacier flow causing increased surface crevassing and hence calving (Benn et al., 2007), or increased fjord circulation (Motyka et al., 2003; Sciascia et al., 2013). However, Schild and Hamilton (2013) emphasize that the timing of these effects are modulated by glacier geometry in individual cases.

CHANGE IN WEST GREENLAND

The glaciers in the NW region show rather constant retreat rates throughout the period with the smaller retreat glaciers showing retreat slowing after 2005 (Fig. 6). In common with other areas, the NW sector showed large losses during 2004 and 2005, and 2006 was a year when many glaciers in the NW advanced (Fig. 4). GRACE analyses suggest acceleration in mass loss in the NW since late 2005 or 2007 (Chen et al., 2011; Khan et al., 2010; Schrama and Wouters, 2011; Harig and Simons, 2012; Sasgen et al., 2012), with the highest mass loss rates at the end of the period, (i.e., since 2009). In contrast, Enderlin et al. (2014) showed glaciers in the NW having a slowly increasing discharge over the period with a decrease in discharge acceleration during 2006–2008 and an increase in 2008–2010. Our data show retreat was almost ubiquitous in the region over the period. Thus there is no convincing frontal response that would match the GRACE signal acceleration.

McFadden et al. (2011) studied 59 glaciers combining those in the NW with the SW during 2000–2009. Their study showed that while most of these glaciers retreated, the majority retreated less than 1 km and that retreats were asynchronous across the region. Our data show that between 2000 and 2010 most glaciers in the NW (59 out of 91) retreated more than 1.5 km, whereas most glaciers (13 from 22) in the SW retreated less than 1.5 km (Figs. 1 and 6). McFadden et al. (2011) also analyzed the relationship of retreat with SST and in accordance with our larger study found no relationship between retreat rates and this variable.

The NW region is dominated by marine-terminating glaciers (McFadden et al., 2011). Moon et al. (2012) reported that one-third of these glaciers increased in speed throughout 2000–2010, one-third showed no trend, and a quarter slowed: a further 15% of the region's glaciers slowed in 2000–2005 and then accelerated substantially. This complex pattern is summarized by Moon et al. (2012) as a general trend of speedup throughout 2000–2010, with the rate increasing toward the end of the period, especially during 2007–2010.

Our data confirm that there is no strong and synchronous signal in frontal position from the glaciers, which show a rather constant retreat through the period. There is an indication in the retreats of the smaller glaciers of a slowdown in retreat rate (Fig. 6) starting in 2005–2006 as reported by Enderlin et al. (2014). There is no strong indication of a strong increase in retreat rates. Overall, we suggest that at least a component of the clear GRACE signal of accelerated mass loss results from changes in surface mass balance or thinning at higher elevations (e.g., Howat et al., 2008b) rather than simply reflecting dynamic loss. This conclusion is supported by the study by Sasgen et al. (2012), which found that part of the acceleration in mass loss seen by GRACE in the NW was a consequence of higher precipitation prior to 2005 and lower precipitation after that date.

FIGURE 4.

Annual changes in calving front position. Note that symbols are plotted with the largest symbols lowest and smallest symbols on top, meaning that there are no fully obscured symbols in the figure. The advances in blue are plotted on top of retreats (in red) where they fall in the same size category.

FIGURE 5.

(a) Summed changes over the period 2000–2010 for each sector. (b) Annual contributions of each sector with the brown line following the overall change for each year. A line graph version of part b using the same color scheme is included as Figure 7, part a.

Conclusions

Greenland's marine-terminating glaciers are an important interface between ice and ocean, and their frontal position acts as an indicator for dynamic thinning of the Greenland ice sheet: retreat indicates increased calving flux corresponding to an increased contribution to sea-level rise. We have measured the front positions of 199 Greenland marine-terminating glaciers located in all regions around Greenland over the period 2000–2010 using Landsat satellite data. All regions of the ice sheet are affected by sustained and substantial glacier retreat during this period (Fig. 3). Over the whole period, the 199 glaciers retreated more than 267 km in total, with only 11 glaciers (5.5%) showing overall advance. Only one year of the 11-year record, 2006, was characterized by overall minor advance (Fig. 5). There were considerable inter-annual and inter-regional differences (Fig. 4). In general, the pattern of mass loss detected by GRACE and other measurements is reflected in the calving record of Greenland glaciers.

Glaciers in the SE sector of Greenland show very high retreat rates (-325 m a^-1) during the period 1999–2005, and subsequent stabilization and readvance (+161 m a^-1) during following years: glaciers in the SW show a similar pattern of pre-2005 retreat and subsequent stabilization. In contrast, outlets in the NE show mean retreat of 300 m over the entire 11-year period, with stabilization of glaciers with large retreat from 2005 but those with smaller retreats showing slow retreat punctuated with clear annual variations over the entire period. In the NW, outlets show continuous retreat (-83 m a^-1), with seasonal fluctuations. Our results suggest several regions in the south and east of the ice sheet are synchronized in their behavior and thus likely share controls on their dynamic changes: this is probably oceanic in origin. In the case of the SE, we show a significant correlation between ocean temperature lagged by one year and the frontal position of the glaciers.

FIGURE 6.

Calving front location of outlet glaciers separated on the basis of the range of advance/retreat exhibited by region. Glaciers that retreat less than 1500 m in blue and those that retreat more than 1500 m in red during the period 2000–2010 (the regions and location of glaciers is shown in Fig. 1). Note that the scales for the two classes of glaciers are different (smaller retreats on the left-hand axis; larger retreats on the right-hand axis) and that the scales vary among regions. Lines show monthly running mean (solid line) and median (dashed line).

FIGURE 7.

Different data sets for comparison, (a) Glacier annual frontal position change for each region. (b) Mean annual range anomalies, which show how the mean annual range (maximum — minimum frontal position for each glacier) in each region differs from the long-term average range. (c) MODIS-derived sea surface temperatures (SSTs) (Brown and Minnett, 1999) and temperature for depth range of 200–500 m at the westernmost station of the Faxaflói-line. (d) Melt day anomalies from Greenland Daily Surface Melt 25 km EASE-Grid 2.0 Climate Data Record (Mote, 2012). (e) Coastal sea ice coverage from the Sea Ice Index (Fetterer et al., 2002). SST and sea ice coverage were calculated for the entire continental shelf area for each region.

The record of behavior of the fronts of tidewater glaciers we present has the potential to be used in future studies of controls on calving losses from the Greenland ice sheet. Furthermore, our data could be used to assess the impact of assuming a constant groundling line or front position for studies of ice sheet mass balance that use the surface mass balance minus discharge method. In order to make these sorts of studies as simple as possible, we provide the data as  supplementary material (.kml file S2) (427-447 pp AAAR-47-3-02 supplementary file A1.kmz) to this paper. Understanding the nature, distribution, and controls on dynamic change of Greenland's tidewater terminating glaciers are essential for predicting Greenland's future sea-level contribution.