Reconstruction of GrIS Elevation Change.

As part of NASA’s Program for Arctic Regional Climate Assessment (PARCA), airborne laser altimetry surveys began in 1993 with NASA’s Airborne Topographic Mapper (ATM) (27). However, investigations of ice sheet mass balance and related sea level rise were hampered by the lack of spatially comprehensive elevation time series. To remedy this, NASA launched the Ice, Cloud and land Elevation Satellite (ICESat) mission in 2003 with the primary goal of measuring elevation changes over the polar ice sheets with sufficient accuracy to assess their impact on global sea level (28). After a successful period of obtaining accurate elevations of the Greenland and Antarctic ice sheets, ICESat’s last campaign ended on October 11, 2009. The successor, ICESat-2, is expected to be launched in 2017. To “bridge” the intervening time without satellite laser altimetry data, NASA started Operation IceBridge mission (OIB), which has been gathering laser altimetry data using the ATM and the Land, Vegetation and Ice Sensor [LVIS (29)] airborne systems in both polar regions.

We developed the novel Surface Elevation Reconstruction and Change detection (SERAC) method to determine surface elevation changes at ICESat crossover areas (intersections of ascending and descending ICESat tracks) (30). The method is based on fitting an analytical function to the laser points of a surface patch, such as a crossover area, of ∼1 km² in size. The surface patches of different time epochs at the same crossover area are related to each other; we have introduced the constraint that within a surface patch, the shape of the ice sheet remains the same over the entire observation period; only its absolute elevation changes. The least-squares adjustment of SERAC simultaneously determines one set of best-fit shape parameters and a time series of elevations for all time epochs involved, together with rigorous error estimates (ref. 30, SI Text, and Fig. S1).

Originally limited to ICESat crossover areas only, SERAC has been extended to provide solutions along the ICESat ground tracks by combining ICESat data with airborne laser altimetry data (31). In this way, the spatial density of surface elevation time series increases dramatically, as Fig. 1 vividly demonstrates. Fig. 1B depicts ICESat crossover locations (brown) and additional locations where 2003–2009 ATM and LVIS flights intersected or repeated ICESat ground tracks (blue). However, large gaps remain, especially in southern Greenland. Adding ATM data from the period 1993–2002 as well as ATM and LVIS data from 2010 to 2012 remedies this situation, resulting in a dense data set.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Location of GrIS elevation time series according to the main data sets used for SERAC reconstruction. (A) PARCA ATM (1993−2003) and ICESat (purple); (B) PARCA ATM (2003−2009) and ICESat (blue), ICESat crossovers only (brown); (C) OIB ATM/LVIS (2009−2012) and ICESat (red); and (D) combined: all solutions (black). GrIS is shown in gray, and land surface with local ice caps and glaciers is in green/brown hues. Symbols in D mark the locations of elevation change time series shown in Fig. S1.

The fusion framework of SERAC offers other advantages. For example, inclusion of ATM and/or LVIS data that were collected during ICESat mission (2003−2009) increases the temporal resolution of the elevation change record. If data are available from earlier time periods, the time series are extended backward in time, before 2003. Using data from the OIB mission extends the time series toward the present.

Ultimately, by combining all NASA laser altimetry measurements, elevation time series are reconstructed at ∼100,000 locations, resulting in a very dense coverage along ICESat ground tracks, especially in the ice sheet marginal region. Despite occasional cloud cover, ice sheet elevations were measured at least once during each of the 19 ICESat operational periods at most crossover locations (30). Thus, by adding ATM and LVIS measurements, a dense temporal sampling is obtained for 2003–2009, with additional points from LVIS and ATM extending most of the curves beyond ICESat’s lifetime (Fig. 2A and Fig. S1). After removing the effect of vertical crustal motion due to Glacial Isostatic Adjustment (GIA; SI Text), we partition the ice thickness change time series into components associated with ice dynamics and SMB changes (Materials and Methods, SI Text, and Fig. S1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Classification of outlet glaciers based on dynamic thickness change pattern. (A) Thickness change time series derived from the combined ICESat/ATM/LVIS altimetry record (1993−2012) illustrating different dynamic outlet glacier behaviors. Thickening: Store Glacier (6); no dynamic change: Petermann Glacier (78); decelerating thinning: Kjer Glacier (83); accelerating thinning: Zachariæ Isstrom (51) and Ikertivaq NN (46); full cycle thinning: Jakobshavn Isbræ (1); thinning with varying rate: Midgård Glacier (121); thinning, thickening, and thinning with abrupt termination of initial thinning: Helheim (3), Koge Bugt C (4), and A. P. Bernstorff (12) glaciers; unique pattern with periodic thinning and thickening: Daugaard-Jensen Glacier (8). Numbers in parentheses are ID numbers from ref. 32 and in Table S1. Gray box marks the duration of the ICESat mission, and glacier locations are shown in B. Dynamic thickness changes of the four large outlet glaciers, Jakobshavn Isbræ, Kangerlussuaq, Helheim, and Petermann glaciers, underlined in the figure, are modeled in refs. 9⇓–11. (B) Distribution of different outlet glacier behavior types over a background of ice sheet bed elevation from ref. 45. Inset shows the detailed pattern north of Jakobshavn Isbræ overlain on ice velocities from ref. 32. Abbreviations mark the following outlet glaciers: Sermeq Avannarleq (SA, 53), Sermeq Kujalleq (SK, 13), Kangilerngata Sermia (KS, 52), Eqip Sermia (ES, 90), Kangiata Nunaata Sermia (KNS, 36), Skinfaxe (S, 82), Rimfaxe (R, 58), and Heimdal (H, 39) glaciers. See Table S1 for a complete list of glaciers and their classification based on 2003–2009 and 1993–2012 dynamic thickness change patterns.

The high spatial density of the new 1993–2012 elevation change record and the 91-d repeat cycle of ICESat allow for the investigation of the spatiotemporal pattern of ice sheet thickness change at different scales. The ice thickness change time series (Fig. 2A and Fig. S1) provides the highest resolution, suitable for characterizing the dynamic processes affecting individual outlet glaciers. The most recent compilation of GrIS ice velocities includes 242 outlet glaciers with a width greater than 1.5 km (32). We identified 130 of these glaciers with a 5- to 19-y-long altimetry record, out of which 115 are marine terminating (Tables S1 and S2). Average elevation change rates are found to be in good agreement with previous studies (SI Text and Table S3). However, we have shown that changes are typically nonlinear in time and most of the rapid changes occur during the ICESat mission (Fig. 2A). For many marine-terminating outlet glaciers, dynamic thickness change patterns are consistent with an inland propagation of dynamic thinning or thickening initiated at the coast (Fig. S2 A and B). Some glaciers exhibit a more complicated behavior, however. For example, Størstrommen, L. Bistrup Bræ, and Marie Sophie glaciers, which are quiescent surging glaciers, have a characteristic pattern with large, steady thickening at their source regions, as ice accumulates upstream of the reduced flow, and thinning below the area where the surge was initiated (Fig. S2C and Table S1), while the complex elevation change pattern of Hagen Bræ might indicate an ongoing surge (Table S1). Short-term, sometimes cyclic elevation changes occurred on 15 outlet glaciers, all marine terminating (SI Text and Table S1), and may indicate control from subglacial hydrology or are perhaps related to the drainage of proglacial lakes (e.g., Daugaard-Jensen Glacier, Fig. S2D). Dynamic thinning was negligible on 13 out of the 15 land-terminating glaciers (SI Text and Tables S1 and S2). To facilitate interpretation, glaciers are divided into the following distinct groups according to their dynamic thickness change pattern in 2003–2009: thinning with steady or slowly changing rates (accelerating, decelerating, full cycle thinning); slow or rapid thinning that abruptly terminated and was followed by thickening and in some cases by resumed thinning; thickening; unique elevation change pattern; and no dynamic change (Fig. 2 and Table S1).

To investigate drainage basin-scale processes, we compute annual ice thickness change rates at each surface patch location from a polynomial fit through the thickness changes reconstructed by SERAC (30) and partition these rates into changes associated with SMB and ice dynamics (Materials and Methods, SI Text, and Fig. S1). The interpolated annual thickness change rate grids show intricate and rapidly changing patterns (Fig. 3 and Movie S1). To quantify these, volume and mass change rates of the main ice sheet regions are calculated (Fig. 4 and Tables S4 and S5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Annual total, SMB-related, and ice dynamics-related thickness change rates of the GrIS for 2003–2009 balance years from ICESat, ATM and LVIS laser altimetry observations (see Fig. 1 for locations of elevation change records). Dotted lines on the dynamic thickness change maps mark the ELA (average 2003–2009 SMB = 0). Ice sheet boundary is from ref. 46, and black regions show weakly or not connected glaciers and ice caps. Balance years start on September 1 and end on August 31 of the following year. See Movie S1 for a higher-resolution, animated version of the figure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Annual mass change rates in gigatons per year for major drainage basins shown in Fig. 2B. Annual total mass change rates from laser altimetry (red) are partitioned into mass changes due to SMB (blue) and ice dynamics (green). Annual mass change rates and their error estimates are listed in Table S5.

GrIS Ice Thickness Change and Mass Loss Patterns, 1993–2012.

We estimated a mean annual GrIS loss of 243 ± 18 Gt⋅y⁻¹ (277 ± 7 km³⋅y⁻¹), equivalent to 0.68 ± 0.05 mm⋅y⁻¹ SLR for 2003–2009 (Tables S4 and S5). This mass loss and its interannual variability are in good agreement with the reconciled GrIS mass loss estimate derived from a combined ensemble of laser altimetry, GRACE, and IOM data (3). However, we detected higher average mass loss and interannual variability than the laser altimetry results included in ref. 3, bringing the laser altimetry, GRACE, and IOM estimates closer to each other (ref. 3, SI Text, and Tables S5 and S6), thus reconciling the previously perceived inconsistencies among different methods. Dynamic thinning contributed 48% to the total mass loss, which is the same as reported in ref. 4. Dynamic loss was largest in 2004–2005, followed by a gradual decrease that was balanced by accelerating SMB loss (Fig. 4). However, at the same time, the relative contributions of major drainage basins changed significantly (Figs. 3 and 4), indicating that processes acting on time scales of less than a decade have a significant effect on ice sheet mass loss and related SLR. The only region exhibiting steady mass loss from 2003 to 2009 was Jakobshavn, while mass loss decelerated from southeast and east Greenland and accelerated from the rest of Greenland, confirming the pattern reconstructed from GRACE observations (23). In this section, we review the dynamic behavior of individual marine-terminating outlet glaciers and explore their impact on drainage basin scale dynamic mass changes.

Almost half of the total 2003–2009 GrIS mass loss originated from southeast Greenland (Table S5). Glaciers in this sector thinned rapidly between 2003 and 2005, reaching their peak discharge in 2005 (refs. 16 and 18, Figs. 3 and 4). During this time, dynamic thinning extended far inland, up to the ice divide in some areas (Movie S1), maintaining a pattern that started in the 1970s or earlier (33). This large dynamic loss, peaking at 166 ± 31 Gt⋅y⁻¹, was a major source of the record negative GrIS mass balance of 293 ± 38 Gt⋅y⁻¹ in 2004–2005. As flow acceleration slowed down or reversed to deceleration after 2005 (16), thinning rates decreased, and several glaciers, for example Helheim Glacier, started to thicken (Fig. 2). At the same time, the long-term trend of high-elevation thinning also reversed, and by 2007–2008, most of southeast GrIS exhibited dynamic thickening (Fig. 3 and Movie S1). By 2007–2008, the ice loss rate from the southeast GrIS dropped to less than one third of its peak value, as a result of a diminishing dynamic mass loss. However, the slowdown and thickening of outlet glaciers was short lived, as they resumed acceleration (16) and thinning by 2009 (Fig. 2). In addition to widespread short-term changes, outlet glacier thinning shows a large spatial variability in southeast Greenland, much like the velocity record (16), indicating an intricate interplay of regional and local forcings and controls. For example, rapid thickening of outlet glaciers within a region extending to 500 km in north−south direction, and including Helheim, Køge Bugt C, and A. P. Bernstorff glaciers, started at the same time and exhibited very similar patterns (Fig. 2), suggesting regional climate controls. Meanwhile, other glaciers in the region, such as Midgård and Ikertivaq glaciers, continued to thin, losing ice at increasing rates. In contrast, Heimdal, Rimfaxe, and Skinfaxe glaciers, maintaining steady calving front positions since 1933 (34), have been thickening (Fig. 2B).

Concurrent with the 2003–2005 rapid thinning of the southeast region, the adjacent southwest basin was thickening (Fig. 3 and Movie S1). This positive mass balance was due to the dynamic thickening of the land-terminating ice sheet margin, interpreted as a reaction to increasing accumulation during an ice sheet readvance 4,000 y ago (35). Increasing dynamic thinning of major outlet glaciers (e.g., Kangiata Nunata Sermia, Fig. 2B) and accelerating SMB loss resulted in an overall negative mass balance of this region by 2005 (Fig. 4).

Annual mass loss of the Jakobshavn region was steady at a rate of 30 ± 4 Gt⋅y⁻¹, dominated by losses caused by the continuing speedup and corresponding thinning of Jakobshavn Isbræ (19). Thinning rates started to decrease near its calving front in 2007 (Fig. 2A and Fig. S2A), indicating an adjustment to new environmental conditions and signaling a potential future mass loss decrease. The outlet glaciers draining to the narrow fjords north of Jakobshavn Isbræ show a complex spatial pattern of dynamic elevation changes in 2003–2009 (Fig. 2B, Inset).

In northwest Greenland, ice loss has accelerated linearly from 31 ± 11 Gt⋅y⁻¹ to 83 ± 18 Gt⋅y⁻¹ between 2003 and 2009, due to increasingly negative SMB anomalies and a steady dynamic loss (Fig. 4 and Table S5). Our long-term altimetry record shows that dynamic thinning has been steady or accelerated on most outlet glaciers during the last 15–20 y (e.g., Kjer Glacier, Fig. 2). This is consistent with the steady increase of ice discharge between 2000 and 2010 detected by refs. 16 and 36 and contradicts a previous reconstruction that indicated a stable period between 1992 and 2005, followed by dynamic thinning and increased discharge (37).

The three other major regions (north, east, and northeast) remained dynamically relatively inactive over the period of 2003–2009. Ice sheet mass balance had a similar trend in north and northeast, where a decreasing negative balance was followed by a slow mass loss increase since 2005 due to a combination of increasing negative SMB and increasing dynamic loss. Thinning rates of north and northeast Greenland outlet glaciers are relatively small (<5 m⋅y⁻¹). However, thinning of Ryder Glacier and Zachariæ Isstrom (Fig. 2) at current or increasing rates could unground their large ice plains within a few decades as continuing thinning brings the ice closer to flotation (21, 26). The resulting speedup over large areas would ultimately cause a significant mass loss from the deep central part of the GrIS. Elevation changes were also small, but increasingly positive, in east Greenland, resulting in a positive mass balance by 2007.