Ice Sheets and Ice Sheet Mass Balance 1

The supraglacial lakes and streams that appear on the surface of the Greenland ice sheet during summer are one component of the dynamic and complex hydrological system of the Greenland ice sheet. These lakes and streams play a major role in the transport and detention of surface meltwater as well as in the evolution of the englacial and subglacial channels which, in turn, impact ice velocities and ice dynamics. In recent years, several studies have focused on quantifying lake area, depth, and timing of their formation and disappearance, using spaceborne visible and near-infrared data at different spatial and temporal resolutions. Moreover, recent studies using high-resolution commercial satellite imagery have been attempting to quantify the spatial and temporal resolution of supraglacial streams and rivers.

Still, despite its importance, little is known about the surface hydrological system during winter, with the only examination to date of the potential over-wintering of the lakes being conducted by Koenig et al. [2015], with this work based on the analysis of data from the Operation IceBridge.

Here, we use a combination of data from the recently launched Sentinel-1, WorldView3 and LANDSAT8 satellites to study the evolution of supraglacial lakes and streams during both summer and winter over three study sites of the Greenland ice sheet. The sites are located nearby the Ryder, Petermann and Jakobshavn glaciers. Water bodies in summer and spring are identified based on a multi-spectral approach using Landsat8 and WorldView3 data and their persistency and evolution during winter is analyzed by means of the Sentinel-1 radar data. The analysis of remote sensing data is complemented with the outputs of a regional climate model simulating the evolution of surface parameters, such as accumulation and runoff, as well as those of an electromagnetic model simulating the recorded Sentinel-1 backscattering values as a function of the physical properties of the over-wintering lakes (e.g., snow depth covering lakes, thickness of the ice layer, etc.). 

Numerous glaciers in Greenland have sped up rapidly and unpredictably during the first part of the 21^st Century. We started the Greenland Ice Mapping Project (GIMP) to produce time series of ice velocity for Greenland’s major outlet glaciers. We are also producing image time series to document the advance and retreat of glacier calving fronts and other changes in ice-sheet geometry (e.g., shrinking ice caps and ice shelves). When the project began, there was no digital elevation model (DEM) with sufficient accuracy and resolution to terrain-correct the SAR-derived products. Thus, we also produced the 30-m GIMP DEM, which, aside from improving our processing, is an important product in its own right.

 Although GIMP focuses on time series, complete spatial coverage for initializing ice-sheet models also is important. There are insufficient data, however, to map the full ice sheet in any year. There is good RADARSAT coverage for many years in the north, but the C-band data decorrelate too quickly to measure velocity in the high accumulation regions of the southeast.  For such regions, ALOS data usually correlate well, but speckle-tracking estimates at L-band are subject to large ionospheric artifacts. Interferometric phase data are far less sensitive to the effect of the ionosphere, but velocity estimates require results from crossing orbits. Thus, to produce a nearly complete mosaic we used data from multiple sensors, beginning with ERS-1/2 data from the mid 1990s. For the faster moving ice-sheet margin where phase data cannot be unwrapped, we used speckle-tracking data. In particular, we have relied on TerraSAR-X for many fast-moving glaciers because the ionosphere far less affects X-band data. This pan-Greenland velocity map as well as many of the time series would not have been possible without an extensive archive of data collected using six satellites from four different space agencies, much of which was distributed through the Alaska Satellite Facility (ASF). The record we are producing is being distributed to the wider community through the National Snow and Ice Data Center (NDISDC) and serves as an important precursor data set for the NASA ISRO Synthetic Aperture Radar (NISAR), launching in 2020.

The AltiKa altimeter on-board SARAL, a joint CNES/ISRO mission, was launched in February 2013 on the same 35-day orbit than previous European altimeters, Envisat and ERS1-2. SARAL/AltiKa is thus, a unique opportunity to extend the repeat observations of this orbit, surveyed since 1991. Moreover, the altimeter operates in Ka-band, a higher frequency than previous ones and offers new fields of investigation. In particular, the theoretical penetration within the snowpack is reduced when compared to Ku-band altimeters which is confirmed by cross-over analysis and along-track survey. Indeed, we show that the impact of backscatter changes on height decreases from 0.3 m/dB for Ku-band to only 0.05 m/dB for Ka-band. We show that the volume echo in Ka-band results from the near subsurface layer and  is mostly controlled by ice grain size, whereas in Ku-band, internal stratifications also play a role. With the help of nadir looking angle radiometer on-board altimetric mission, SARAL/AltiKa offers an opportunity to link backscatter and emissivity with grain size or snowpack characteristics. Finally, we will show results of the extend of the temporal series from ERS-2 and Envisat up to the 3 years of the AltiKa flight period. In particular, we will show the height losses acceleration of outlet glaciers of the Amundsen Sea.

Mass changes of the Greenland ice sheet may be estimated by the Input Output Method (IOM), satellite gravimetry, or via surface elevation change rates (dh/dt). While the first two methods have been shown to agree well in reconstructing ice-sheet wide mass changes over the approximately the last decade (since 2003 for gravimetry), there are few long records from satellite altimetry and none that provide a time evolving trend that can be readily compared with the other methods. Here, we interpolate radar and laser altimetry data between 1995 and 2014 in both space and time to reconstruct the evolving volume changes. We use a novel interpolation approach that incorporates prior information related to ice dynamics. A firn densification model forced by the output of a regional climate model is used to convert volume to mass.

We find that mass changes are dominated by SMB until about 2001, when mass loss rapidly accelerates. The onset of this acceleration is somewhat later, and less gradual, compared to the IOM. Our time averaged mass changes agree well with recently published estimates based on gravimetry, IOM, laser altimetry, and with radar altimetry when merged with airborne data over outlet glaciers. We demonstrate, that with appropriate treatment, satellite radar altimetry can provide reliable estimates of mass trends for the Greenland ice sheet. We also examine the impact of the extreme melt event in 2012 on the volume and, consequent, mass change for that year and compare with GRACE to assess how well our methodology accounts for this anomaly. Our time series could, potentially, be extended back to 1992 with the aid of ERS-1 data but the early part of this mission had varying orbit repeat cycles, which adds complexity to the analysis. The Cryosat-2 data used, since 2010 provides robust results, which can be extended for the length of the mission and supplemented with Sentinel 3 data after its launch.

The Greenland Ice Sheet (GrIS) experienced two years of extreme mass balance in 2012 and 2013; while 2012 exhibited exceptionally vigorous ice loss in summer, the ice sheet was close to balance in 2013. Here we analyse the output of two GrIS surface-mass balance models and compare them to GRACE and CryoSat-2 satellite observations. We show that the 2012 losses were caused by enhanced melt production, mainly originating in the south-west of the ice sheet. In contrast, melt production was much lower in 2013 – making this year similar to the conditions observed between the years 1960-1990, in which the ice sheet is considered to have been close to balance. To a large extent, the observed mass variability is a consequence of the large-scale North Atlantic atmospheric circulation, as reflected by the North Atlantic Oscillation (NAO) index; circulation either favours transporting heat from the mid latitudes along the west coast of the GrIS or allows for more cold air to flow from the Canadian Arctic. We show the level of agreement between GRACE / CryoSat-2 and the atmospheric models, infer the snow, net melt and ice dynamic contributions in the 2012 and 2013 mass balance anomalies, and discuss whether pre-conditioning of the ice sheet in 2012 played a role for the mass balance in 2013. The transition from extreme melt year to a near-balance year is unprecedented in the multi-decadal records of the atmospheric models; and both years are exceptional regarding the trends observed in the last decade.