Range and azimuth shifts are widely used to estimate geophysical motion from SAR images, in particular for glacier velocities. In polar regions, the measurement is often corrupted by spurious shifts induced by ionospheric variations. This happens not only with L-band data but even for C-band, because of the extreme sensitivity of the azimuth shifts to phase errors within the synthetic aperture. The azimuth shifts are in particular proportional to the first azimuth derivative of the ionospheric phase screen. In the C-band example attached to this abstract (Sentinel-1 IW mode, Northern Greenland), one can see errors of about +/-5m. The situation is even worse for lower frequencies bands like L-band.

The first idea to correct for the azimuth shift is to use the range split-spectrum technique, which has proven very effective to remove L-band and C-band ionospheric phase screens [1]. Let us assume that we want to reach 1 m precision in the azimuth shift with Sentinel-1 IW mode. This corresponds to about 1/20 of the azimuth resolution, which means that we need 1 mm (13 deg) precision in the phase errors at the edge of the aperture seen by a given target. The formula for the performance of the split-spectrum technique predicts that we need to average over 1 million pixels, i.e. around (10 km)^2, considering the resolution of Sentinel-1 IW mode. This means that the split-spectrum technique can only correct rather smooth ionospheric disturbances in the azimuth shifts. An example is to be found in the attached figure (first row; left: original shifts, right: after compensation exploiting range split-spectrum). It is clear that we could compensate only large scale components of the disturbance.

To estimate finer scale spatial variations we propose exploiting the difference between the azimuth shifts calculated for a lower and an upper sub-band of the target azimuth bandwidth (the differential shifts). The two shifts are identical for a physical shift on the ground and will cancel out in the difference, but they will differ in case of ionospheric oscillations. In particular the shift difference corresponds to the second azimuth derivative of the ionospheric phase screen, with no contribution from the actual ground motion. Considering that the azimuth shifts reflect the first derivative of the phase screen, the correct operator to retrieve the azimuth shifts from the differential shifts will be roughly an integrator in azimuth direction.

The figures in the second row show the azimuth shifts (left) and the differential shifts (right) for a single burst. It is clear that the differential shifts are rather noisy and an averaging step is strongly needed. We propose to solve the issue of determining the right operator and spatial averaging in one step, with the use of a prediction error filter.
We consider a generic 2D filter (f) to be applied to the differential shifts (d) and we determine it by minimizing the prediction error (PE) with respect to the observed azimuth shifts (s). In vectorial form, we minimize

PE = E[(f^t . d - s)^2]

across the burst (or a significant portion of a burst). The operator "^t" is the transpose operator. The usual least square solution is:

f^t = E[d . d^t]^-1 E[s . d^t],

and the predicted shifts (PS) are given by applying the filter to the differential shifts:

PS = f^t . d

Figures in the third row display the estimated azimuth shifts (left) and the residual shifts (right), which should contain only ground motion plus components of the ionospheric disturbances that the filter was not able to estimate correctly. As expected the differential shifts are able to predict the high-frequency components of the shifts, and leave lower frequency components virtually untouched.

The integration of the split-spectrum and prediction error filter corrections is still on-going work. We are also limited by some artefacts of the S1 processor, which show up as azimuth dependent slopes in the split-spectrum processing (i.e. in the estimated of the ionospheric screens).

Future work will also be directed at analysing the performance of the prediction error filter; we will then be able to tell to which extent it is possible to correct the ionospheric disturbances in the azimuth shifts using a combination of split-spectrum and differential shifts. It is important to know whether or not the two proposed corrections leave a spectral gap between low spatial frequencies (corrected with range split spectrum) and high spatial frequencies (corrected with the azimuth differential shifts).

Without any correction, the use of azimuth shifts in presence of ionospheric disturbances is severely limited.

[1] Toward operational compensation of ionospheric effects in SAR interferograms: The split-spectrum method, G. Gomba et al., IEEE Trans. on Geosci. and Remote Sens., (2016)

Area D - Board 471

SAR images are widely used to measure ground displacement at the Earth's surface such as landslides, earthquakes, volcanoes, glacial flows, etc. Despite the wealth of the available data, spatio-temporal gaps still exist in displacement measurement time series because of either the quality of raw data or the limitations of the displacement extraction method. These measurement gaps hinder a full exploitation of the displacement measurement time series which provides useful information about the spatio-temporal evolution of the phenomenon under observation. Yet there is a substantial lack of studies that are adapted to reconstruct missing data in SAR displacement measurement time series with embedded error analysis. Therefore, we propose a novel data-adaptive method based on the Empirical Orthogonal Function (EOF) analysis of the temporal covariance of a displacement measurement time series to retrieve missing data. This EOF-based method decomposes the temporal covariance in different variability modes, from which coherent spatial patterns whose magnitudes vary over time can be identified. Starting from an appropriate initialization of missing data, this method allows for efficient data reconstruction and denoising by selecting an optimal number of EOFs modes in an iterative process. The choice of the optimal number of EOFs modes relies on a convergence criterion based on the cross validation technique, which aims at minimizing the cross-RMSE, which is the root mean square error between the reconstruction and cross validation data. We evaluate the reconstruction performance, i.e. its ability to find the optimal number of EOFs to reconstruct the signal, on a synthetic displacement field with different complexities in function of Signal-to-Noise Ratio (SNR), quantity and type of missing data and number of iterations. After this validation step, the proposed method is applied to displacement time series containing missing data issued from 1) offset tracking of 65 Sentinel-1 SAR images acquired between October 2016 and December 2017 over Argentière glacier and 2) differential interferometry of 30 Sentinel-1 SAR images acquired between October 2016 and March 2017 over the Gorner and Miage glaciers. Besides the cross validation, the results obtained are also analyzed with residuals and in situ measurements where available to demonstrate the potential of the method to reconstruct incomplete displacement measurement time series of alpine glaciers.

Area D - Board 472

Sentinel-2 imagery, with its repeated acquisition, represents a unique opportunity to quantify spatial and temporal changes of glaciers world-wide. Glacier velocity has been measured from satellite scenes for decades now, yet a range of satellite missions, feature tracking programs, and user approaches has made it a laborious, and sometimes, subjective task. Up-to-date, there has been no tool developed that would allow a user to obtain displacement maps of specified glaciers simply by establishing the temporal and spatial parameters. We present development of a semi-automatic, open-source processing toolbox for retrieval of displacement maps with application to obtaining glacier surface velocities. The toolbox combines the download, pre-processing, feature tracking, and post - processing of highest resolution Sentinel-2A and Sentinel-2B satellite images into an automatic toolbox, leaving a user with a set of rasterized and georeferenced glacier flow magnitude and direction maps for their further analyses. The solution is freely available to everyone in the community and is tailored so that non-glaciologists and people with limited GIS knowledge can benefit from it. The system can be used further to provide the community with a regional and global set of ice velocity change, but can also be modified in a way to detect changes in glacier and snow extent, for instance. Together with the toolbox functionality, we present results of applicability of Sentinel-2 imagery for estimation of ice velocity changes over a range of glaciers, including Jakobshavn Isbræ, Skeidararjökull, Tasman glacier and multiple glaciers in mainland Norway. We discuss the potential and limitations of using Sentinel-2 imagery for tracking ice displacement over time using glaciers of varying size, speed, and subjected to a range of snow and debris cover conditions. Equally, we demonstrate suitability and the limitations of Sentinel-2 imagery for mapping the glaciers in mainland Norway. This work is a contribution to the Copernicus Glacier Service of Norway.

Area D - Board 473

Despite techniques for the creation of glacier outlines have evolved in recent years, mapping debris-covered parts of the glacier is still limited due to its spectral resemblance with surrounding areas in case of optical technique and insulating effect of thick debris cover in case of thermal technique .This leads to immense work load by manually editing the glacier outlines which hampers regular updates of global glacier inventories. Interferometrically derived SAR coherence can be used as an alternative to overcome these limitations. It is a quantitative measure of cross-correlation of the observed surface. The coherence value of 1 signifies the stability, while 0 shows complete temporal decorrelation of the surface.

In Lippl et al. (2018) an automatic processing chain to delineate glacier outlines using InSAR coherence as a primary estimator, and slope as well as morphological operations as secondary filters was developed. Due to availability of scenes we were only able to compare our results for land-terminating stable terminus of Yazgyl Glacier located in the mostly dry Karakoram Range. We compared X- (TerraSAR-X), C- (Sentinel-1) and L- band (ALOS PALSAR) radar frequencies. The impact of different processing settings like resolution, coherence window size and topographic phase removal on the quality of the generated outlines was tested. The intermediate radar frequency C-band data outlined the glacier boundary of Yazgyl Glacier with the least misclassifications and a type II error of 0.47 % compared with Global Land Ice Measurements from Space (GLIMS) data. Therefore, Sentinel-1´s short repeat intervals together with the automatic processing chain can provide a step further for real time monitoring of glacier changes at large scale. Nevertheless, different climatic or topographic conditions with more rainfalls or steeper slopes than for the Gourdon Glacier could complicate or even prevent the processing of accurate outlines.

Therefore, we apply our processing routine for different study sites around the world to test the robustness of our approach with the freely available Sentinel-1 data. We would like to present particular results which are difficult to interpret in this poster to stimulate the discussion within the glaciological/remote-sensing community.

Reference:
Lippl, S., Vijay, S., Braun, M. (2018): Automatic delineation of debris-covered glaciers using InSAR coherence derived from X-, C- and L-band radar data: a case study of Yazgyl Glacier. Journal of Glaciology, 64 (247): 811-821. doi:10.1017/jog.2018.70

Area D - Board 474

In this poster we present velocity fluctuations and area changes of several selected glaciers on James Ross Island, Antarctica, and analyse their correlations with different climate variables.

James Ross Island is located east from the northern Antarctic Peninsula in a region which has undergone considerable changes in the last decades with the retreat or disintegration of several ice shelves. The glaciers of James Ross Island have shown predominantly retreat over the last decades while some glaciers showed also advance.

Area changes were estimated by a detailed analysis of glacier front variations at a temporal resolution of about 2-4 months by using Landsat-8 Panchromatic band. Glacier velocities were derived with intensity feature tracking mainly from TerraSAR-X scenes between 2014 and 2018 with a temporal time step of ~1,5 months and a spatial resolution of 50 m. We validated these remote-sensing-derived data with daily in-situ data from GPS loggers installed on Gourdon Glacier on the east side of the island and found good accordance. Gourdon Glacier showed an acceleration phase in austral summer 2014/2015 which prolongated over the whole winter season until about mid of 2016. We found different velocity patterns for most of the other glaciers on James-Ross-Island. To test if these changes could be driven also by internal parameters without direct climatic influence, we analyse correlations of a set of available climate variables from in-situ data, MAR model, and Merra-2 re-analysis, with the velocity patterns of the individual glaciers. We are testing the correlations also with different time-lags to estimate delay in glacier response.

Furthermore, the goal of our study is to include the measured velocities in a mass change estimate with the flux-gate approach (input-output method) and to compare values with geodetic glacier mass balances from TanDEM-X height models. In order to improve the ice flux estimates we operated a comprehensive airborne ground penetrating radar (GPR) campaign on Gourdon Glacier in February 2017 and 2018. Ice thickness and elevation data were collected on the outlet and the catchment area of Gourdon Glacier while including repetitions of profiles flown by previous ice thickness surveys in order to evaluate potential changes in the accumulation area.

Area D - Board 475

Mapping snow cover (SC) on glaciers at the end of the ablation period provides a possibility to rapidly obtain a proxy for their mass balance. Satellite data provide this information over large regions and back in time and have thus been widely applied for this purpose using a range of satellite sensors. However, as glacier ice originates from compressed snow both have very similar spectral properties and standard methods to map snow from spectral properties struggle to distinguish snow on glacier ice. Consequently, many studies applied manual delineation for mapping the snow extent on glaciers. As this is too time consuming for large regions and dense time series, an automated way of snow mapping on glaciers has to be found. As for the visual interpretation, the best way of distinguishing the often darker ice from the generally much brighter snow is its spectral reflectance. However, in steep mountain topography (where glaciers are located), reflectance from the ground is modified by the direction of sunlight compared to the normal vector of the surface. For regions pointing away from the sun, snow can be as dark as ice in sunlight. It is thus required to perform a topographic correction of the top of atmosphere reflectance (TOAR) received at the satellite sensor with a DEM of sufficient quality. Converting the mapped snow extent into an averaged minimum altitude of the snow line has several further challenges as the lower part of snow cover on a glacier is often patchy and its lower limit not parallel to elevation contours.

In the study we present an automated tool, named ‘ASMAG’, which automatically maps SC and snow line altitude (SLA) on individual glaciers from multi-temporal Landsat satellite images, a DEM, and given glacier extents. The method has been developed on the example of the Ötztal Alps in Austria, where an evaluation is possible using 30 years of field-based observations of the annual equilibrium line altitude (ELA) from three glaciers. The algorithm automatically determines cloud cover and converts raw digital numbers into topographically corrected TOAR. Subsequently, a reflectance threshold for the near-infrared band is automatically selected to map snow on glaciers. The SLA is calculated based on the frequency distribution of elevation bins that have more than 50% SC. The accuracy of the SC mapping is about 90% and the SLA is correct (90% confidence) in more than 80% of all cases. Comparison with manually digitized outlines in other regions (Pamir, Andes) show an even better agreement, indicating that the method can also be applied elsewhere.

Over the full study period (1985-2016) the SC during the summer period decreased by ~ 7% and the mean SLA for all glaciers increased by 54 ±19 m. The maximum SLA shows a similar variability as the measured equilibrium line altitude (ELA), but values are systematically underestimated by about 100 m as the date of minimum snow cover was often missed by several weeks. This difference can be strongly reduced when denser time series become available. We conclude that ASMAG is a robust tool to derive SC and the SLA over large glacier samples, but Landsat data alone are too sparse for accurately retrieve maximum snow line elevations within a given year (for mass balance determination). The forthcoming abundance of Sentinel 2 imagery will likely overcome this restriction and will be tested next.

Area D - Board 476

Time series of annual Minimum Snow/Ice (MSI) cover extent since 2000 are being produced at the Canada Centre for Remote Sensing (CCRS) at 250-m spatial resolution over the Northern circumpolar land area. The product is generated using a specialized data processing chain developed at the CCRS that ingests Level 1 imagery from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Science Data Records (SDR), i.e. L1-equivalent imagery, from the Visible Infrared Imaging Radiometer Suite (VIIRS). The input images are spatially enhanced to 250 m using spatial downscaling technique (Trishchenko et al., 2009; Trishchenko, 2018a). The product is generated over the region centered over Canada (22,800 pixels x19,200 lines) in the Lambert Conformal Conic Projection (LCC) and the Northern Circumpolar region (36,000 pixel x 36,000 lines) in the Lambert Azimuthal Equal Area (LAEA) projection. Processing chain includes several key steps: a) remapping and scene identification; b) 10-day clear-sky compositing and corrections; c) multi-temporal analysis of 10-day composites for April-September time interval leading to seasonal snow/ice probability maps; c) minimal snow/ice extent retrievals from the probability maps (Trishchenko et al., 2016). The summer minimum snow/ice extent is a useful indicator sensitive to regional climate variations (Trishchenko and Wang, 2018). Initial evaluation of MODIS-based and VIIRS-based MSI results revealed very high level of agreement between them (Trishchenko and Ungureanu, 2018).

Time series of the minimum snow/ice extent are analyzed in this study for the thirteen fist-order geographical regions identified in the Randolph Glacier Inventory (RGI 6.0) that overlaps with our study area in the Northern Hemisphere. We compare CCRS-derived year-to-year MSI extent variations with a) the RGI 6.0 baseline, and b) series of land cover maps available the European Space Agency (ESA) Climate Change Initiative (CCI) project. Initial analysis of our results revealed two important facts:
1) Annual time series of CCRS-derived MSI extent is self-consistent product that shows very good correlation with seasonal climate conditions;
2) CCRS-derived MSI extent shows reasonable consistency with RGI 6.0 baseline and can be recommended to Glacier Scientific Community, as a source of annual updates and the first order validation data despite its coarser spatial resolution.

The snow/ice probability maps derived at CCRS are freely available from the Canadian Federal Geospatial Platform (FGP) data archive (Trishchenko, 2018b): http://open.canada.ca/data/en/dataset/808b84a1-6356-4103-a8e9-db46d5c20fcf.

This work is supported through the NRCan Climate Change Geoscience Program (CCGP) and the CCRS activity on Long-Term Satellite Data Records (LTSDR) as part of the project “EO Baseline Data for Cumulative Effects”.

References:
Trishchenko, A.P., Y.Luo, K. Khlopenkov, W.M.Park, S.Wang, 2009: Arctic circumpolar mosaic at 250m spatial resolution for IPY by fusion of MODIS/TERRA land bands B1–B7. Int. J. Remote Sensing. 30, 1635-1641.
Trishchenko, A.P., S. G. Leblanc, S. Wang, J. Li, C. Ungureanu, Y. Luo, K. V. Khlopenkov, F. Fontana, 2016: Variations of annual minimum snow and ice extent over Canada and neighbouring landmass derived from MODIS 250m imagery for 2000-2014 period. Can. J. Remote Sensing. Vol. 42(3), pp. 214-242.
Trishchenko, A.P. and S. Wang, 2018: Variations of climate, surface energy budget and minimum snow/ice extent over Canadian Arctic landmass for 2000-2016. Journal of Climate, vol. 31, no. 3, pp.1155-1172. doi: https://doi.org/10.1175/JCLI-D-17-0198.1.
Trishchenko, A.P., 2018a: Re-projection of VIIRS SDR imagery using concurrent gradient search. IEEE Transaction on Geoscience and Remote Sensing. 56(7), pp. 4016-4024, doi: 10.1109/TGRS.2018.2819825.
Trishchenko, A.P., 2018b: Probability maps of the annual minimum snow and ice (MSI) presence over April,1 to September, 20 period since 2000 derived from MODIS 250m imagery over Canada and neighboring regions.” Available at https://open.canada.ca/data/en/dataset/808b84a1-6356-4103-a8e9-db46d5c20fcf.
Trishchenko, A.P. and C. Ungureanu, 2018: Warm season snow/ice probability maps from MODIS and VIIRS sensors over Canada. Proceedings of the IEEE IGARSS 2018. Valencia. Spain. In press.

Area D - Board 477

TanDEM-X DEM is a global digital elevation model (DEM) released by the German Aerospace Center (DLR) based on X-band SAR acquisition done by the TanDEM-X satellite mission. The feature that distinguishes it among other global elevation models is the high ground spatial resolution of 0.4 arcseconds (translating to 12 m/pixel). At the same time the procedure for its creation involves the combination of several DEMs from different TanDEM-X acquisitions spread between 2011 and 2014, which casts doubt on its value for precise glaciological change detection studies.

In this work, we use TanDEM-X DEM in conjunction with several other DEM products. Aerial laser scanning (ALS)-based dataset from April 2013 serves as the ground-truth reference, while ASTER V003 DEM (09.04.2003) and SRTM v3 DEM (February 2000) are used as representation of past glacier state. The comparisons are made to assess the quality of TanDEM-X DEM for glaciological applications and to provide estimates of the geodetic mass balance for Universidad Glacier, located in central Chile.

We use an earlier method of universal sub-pixel coregistration of DEMs to determine the geometric accuracy of the product. Interactions between topography and the radar illumination, such as radar shadowing, produce artifacts and errors in radar-derived elevation products. We propose a slope-aspect based workflow to remove these errors resulting from radar shadowing over steep terrain.

Subtraction of DEMs from different times and sources serves as a basic, yet effective method of analysis of elevation models. Elevation difference maps relative to ALS DEM are averaged and analyzed to obtain accuracy assessment of TanDEM-X DEM, while multitemporal difference maps allowed for glacier mass balance reconstructions.

We show that TanDEM-X DEM meets its stated HRTI-3 and DTED-3 accuracy standards on flat areas around the Universidad glacier, but fails to do so on steep, rugged mountain slopes surrounding the glacier. The vertical accuracy (+/- standard deviation) of the dataset over non-glaciated area is 0.02 +/- 3.48 m. Compared to 0.22 +/- 8.90 of ASTER and -0.11 +/- 7.31 of SRTM it shows TanDEM-X DEMs outstanding quality. Nevertheless, steep areas (with slopes steeper than 40 degrees) introduce large errors to the results and filtering of these values is required for reliable results. In addition, a coarser, 30 m/pixel, version of TanDEM-X DEM performs worse, than the finer product. However, at -0.08 +/- 7.57 it is still better than the two older datasets investigated.

The mean elevation difference between TanDEM-X DEM and ALS reference is only -0.04 +/- 0.45 on Universidad glacier. This means that the dataset can be used as an approximation of April 2013 glacier topographical surface in glaciological studies. The expected error sources – contribution of acquisitions from many years, penetration of radar signal into dry snow in winter – have little impact on the accuracy of the 2013 glacier geometry, thanks to good production procedure. The high, error-prone slopes are also rare on the glacier, thus the aforementioned errors have not influenced the geodetic mass balance results.

Universidad Glacier has been losing mass at a rate of 0.44 +/- 0.08 m of water equivalent per year between 2000 and 2013 as calculated with ALS-SRTM difference map. This value is in general agreement with previously reported mass balance estimated with the glaciological method for 2012-2014. The glacier tongue is thinning at the fastest rate, regardless of the presence of a debris cover. In the accumulation area only very limited areas of positive mass balance are seen, which indicates that the lowering of ice surface due to downwasting exceeds any increase in ice thickness due to accumulation of snow. More research, both in the field and long-term studies with use of remote sensing, is required to ascertain whether this 13-year rate is stable, or if the thinning accelerates with the gradually warming climate of the area.

We present TanDEM-X DEM as a high-quality product ready for use in glaciological change detection studies. The high resolution, 12 m/pixel product should be used preferably to the coarser, 30 m/pixel dataset in change detection studies due to its better accuracy.

Area D - Board 478

Peripheral glaciers on Greenland contribute strongly to global sea-level rise and their volume changes have been have been followed from altimetry sensors such as ICESat and Cryosat-2. However, little is known about their area changes and up-to-date outlines are missing for most regions. The region to the northwest of Jacobshavn Isbrae (Disko Island, Nuussuaq and Svartenhuk Peninsula) in western central Greenland is hosting nearly 3000 glaciers and ice caps of which many are known as surging. This high abundance has to be recognized when interpreting length, area or volume changes in climatic terms. The largest one (Kuannersuit) had a massive surge in 1995/96 with a frontal advance of more than 10 km. A previous study derived a glacier inventory for 2001 from 3 Landsat ETM+ scenes and determined area changes since their assumed LIA maximum extent as mapped from well visible trim-lines (Citterio et al. 2009). With the now available Landsat 8 and Sentinel-2 images combined with DEMs acquired in 1985 (AeroDEM) and around 2013 (ArcticDEM, TanDEM-X), glacier changes (area, volume) over the past three decades can be studied in detail.

In this study we present an updated glacier inventory for 2015/16 derived from Landsat-8 OLI with the band ratio method and subsequent manual editing of debris cover, water, shadow, some seasonal snow, sea ice and icebergs. Sentinel-2 scenes were used to correct debris-covered glaciers as they had less shadow and better visibility of details. In several regions the latter resulted in a different interpretation of these regions compared to the previous inventory. As the ArcticDEM was spatially incomplete, we used the TanDEM-X DEM to derive improved drainage divides from a watershed algorithm as well s topographic information for each glacier entity. The AeroDEM was co-registered to the TanDEM-X DEM before subtraction. Apart from real elevation changes, the difference grid also revealed numerous interpolation artefacts and outliers that restricted the calculation of changes to selected glaciers.

In total, we mapped 2880 glaciers >0.02 km² covering an area of 4720 km². Of these, 73% are <1 km² and cover 10% of the area whereas 8% are >5 km² and cover 65% of the area. Mean glacier elevation is at 1200 m with a strong dependency on aspect and distance from the coast. Since 1985 (2001) glaciers lost about 40% (20%) of their area, resulting in a mean annual area loss rate of -1.4%, i.e. as high as in the Alps. As the 1985 and 2001 inventories have likely included some seasonal snow, the ‘real’ area loss rate is less negative. Strong area losses are found for smaller ice caps at low elevations but several glaciers were also about stationary over the full period or even advanced (or surged) recently. Elevation changes are very heterogeneous with strong signals (about ±150 m) for glaciers that surged during the 1985 to 2013 period (mass gain / loss at lower / higher elevations) or before 1985 (with the reverse pattern). A general trend of increasingly higher snow lines at the end of the ablation period can be observed from 1985 to 2016. Many ice caps and smaller glaciers are now getting completely snow free over the ablation period, resulting in surface lowering also at their highest elevations.

Area D - Board 479

We present an updated, spatially resolved estimate of 2003-2008 glacier volume changes for entire High Mountain Asia (HMA) from ICESat laser altimetry data. The results reveal a diverse pattern that is driven by spatially greatly varying glacier sensitivity, in particular to precipitation availability and changes. We introduce a spatially resolved zonation where ICESat samples are grouped into units of similar glacier behaviour, glacier type, and topographic settings. In several regions, our new zonation reveals local differences and anomalies that have not been described previously. A step-increase in precipitation around 1997-2000 on the Tibetan Plateau (TP) caused thickening of glaciers in the Eastern Pamirs, Kunlun Shan and central TP by 0.1–0.7 m/yr . The thickening anomaly has a crisp boundary in the Eastern Pamir that continues just north of the central Karakoram. Glaciers in the south and east of the TP were thinning, with increasing rates towards southeast. The precipitation increase is reflected by growth of endorheic lakes in particular in the northern and eastern TP. We estimate lake volume changes through a combination of repeat lake extents from Landsat data and shoreline elevations from ICESat and the SRTM DEM for over 1300 lakes. The rise in water volume contained in the lakes corresponds to 4–25 mm/yr, when distributed over entire catchments, for the areas where we see glacier thickening. The precipitation increase is also visible in sparse in-situ measurements and MERRA-2 climate reanalysis data, but less well in ERA Interim reanalysis data. Considering evaporation loss, the difference between average annual precipitation during the 1990s and 2000s suggested by these datasets is 34–100 mm/yr, depending on region, which can fully explain both lake growth, and glacier thickening (Kunlun Shan) or glacier geometry changes (eastern TP). The precipitation increase reflected in these glacier changes possibly extended to the northern slopes of the Tarim Basin, where glaciers were nearly in balance in 2003-2008. Along the entire Himalaya, glaciers on the first orographic ridge, which are exposed to abundant precipitation, are thinning less than glaciers in the dryer climate of the inner ranges. Thinning rates in the Tien Shan vary spatially but are rather stronger than in other parts of HMA.

Area D - Board 480

Glaciers in Andes mountains are predominately retreating and thinning during the last decades due to the climate change. Central Andes of Chile (32-36°S) contains ~4% of the country´s glaciers and ~60 % of inhabitants (~10 million). Therefore, this is of particular importance because, glaciers represent a substantial water supply for agriculture, human consumption, and services.
Precipitation in this area is highly variable with marked interannual variability due to El Niño Southern Oscillation (ENSO). However, between 2010 to 2015 the region has been affected by a severe drought, recorded as unprecedented due the protracted-on time. During this period several glaciers have increased the thinning rates, however, no precise quantification has been made.
High-mountain environments with glaciers are very sensitive to climate changes, this condition prone to generate new proglacial lakes pose a significant risk to downstream communities and can cause serious damage to communities and infrastructure. We focus this research in the El Morado Glacier (EMG), where a proglacial lake has appeared after ⁓1960. Our dataset compresses aerial and historic photographs and satellite image from 1955 to 2018, two precise digital elevation models (DEMs) from Terrestrial Light Detecting and Ranging (LiDAR) Scanning (TLS), and TerraSAR-add-on- for Digital Elevation Measurement -TanDEM-X- data, and lake bathymetry. The TanDEM-X data was processed using differential interferometry using SRTM C-band DEM as reference. Differences resulting from X- and C-band penetration are considered comparing X- and C-band SRTM data. All DEMs were laterally and vertically co-registered to each other. In addition, we reconstruct by a minimal model the annual surface mass balance of EMG using climate data to explore the signals and sensitivity of glaciers to the megadrought through the long-term annual mass balance.
Here we present a complete quantification about the evolution of the proglacial lake which has duplicated the size after 2010, preliminary results indicate a negative glacier elevation change of 2.52 ± 0.05 m a-1 from 2015 to 2018.

Area D - Board 481

The Tibetan Plateau (TP), known as “water tower of Asia”, has a number of mountain glaciers. Previous studies have showed that glacier thickness changes are different in different glaciers as well as in different regions of the same glacier. Analysis of the causes of such spatial patterns can help better understand the hydrological processes in glacial regions. The surface energy balance of a glacier is the main driving factor for thickness change. Glacier surface temperature and surface albedo play a crucial role in glacier surface energy balance. The spatial relationships between glacier thickness change and glacier surface albedo, glacier surface temperature during 2000-2013 and 2013-2017 were analyzed. The purpose of this research is to provide an application reference for Sentinel candidate mission on Land Surface Tempertaure Monitoring (LSTM) at high spatial resolution.
In this study, we focused on four glaciers – one is called Zhadang glacier (subcontinental type glacier) located in the western Nyainqentanglha Mountains, another three are called Parlung No.4, Yalung and Xueyougu (maritime type glacier), respectively, located in the eastern end of the Nyainqentanglha Mountains in south-eastern TP. The thickness changes of the four glaciers were calculated through the topography difference method. Shuttle Radar Topographic Mission (SRTM) Digital Elevation Model (DEM) was used as glacier surface elevation in 2000, while glacier elevations in 2013 and 2017 were calculated by photogrammetry technique using high spatial resolution stereo images acquired by the the imager onboard the ZiYuan-3 (ZY-3) satellite . Glacier surface temperature and surface albedo of the same period over the same glaciers were derived from Landsat 5, 6 and 8 data.
The results from the four glaciers showed that the thickness change was clearly related to surface albedo and surface temperature during these two periods. The relationship between glacier thickness change and glacier surface temperature is non-linear, i.e. the glacier thickness decreased slightly with glacier surface temperature in the low surface temperature zone, and decreased rapidly with surface temperature in the moderate surface temperature zone, while these relationships are specifically significant in the high surface temperature zone. Glacier thickness change had an obvious negative correlation with glacier surface albedo in the glacier ablation zone, but there was no distinct relationship in the accumulation zone. These results imply that the glacier thickness change pattern is strongly related to surface albedo and surface temperature in the above listed glaciers. In addition, Sentinel LSTM will undoubtedly accelerate relevant researches in the future.

Area D - Board 482

Over the last decades mountain glaciers outside the polar regions were strongly affected by global warming. An accelerated loss of glaciated area and surface lowering have been observed by numerous studies from various mountain regions, e.g. the South American Andes, the Himalaya-Karakoram range and the European Alps, resulting in a significant contribution to global sea level rise. Furthermore, serving as an important source of fresh water for agriculture, irrigation, hydro power and also drinking water, particularly in arid and semiarid regions, large-scale monitoring of mountain glaciers becomes an important task for future water management and prediction of water shortages. However, the accuracy and intercomparability of global glacier dynamics is hampered by a broad variety of different measurement methods and a lack of observations in remote areas. Our aim is to provide a comprehensive overview of glacier elevation, volume and mass change and its contribution to sea level rise for the entire glacierized area from 60°N to 56°S by applying a homogenous remote sensing approach. We estimate glacier dynamics over the period 2000 to 2012/13 from space-borne inteferometric synthetic aperture radar (InSAR) data by calculating changes in glacier surface elevation. Therefore we differentiate vertically and horizontally co-registered TanDEM-X elevation models (DEMs) and SRTM C-Band DEM to derive glacier mass changes and the respective contribution to sea level rise. First results were already achieved for the South American Andes and the Asian Himalaya-Karakoram. For South America we covered a glaciated area of approximately 27,000 km² (~ 85% of total glacier area) and found a mass loss of 19.48±0.65 Gt/a (0.054±0.002 mm/a SLE) which was mainly driven by melt processes of the Northern and Southern Patagonian Icefields and particularly the retreat of large outlet glaciers. As another region being subject to on-going glacier change we estimate an overall loss rate of 13.46 Gt/a for a glaciated area of 80,000 km² based on preliminary results in the High Asia mountain ranges. Currently our measurements are extended by elevation change estimates of glaciers in the Alps, North America and New Zealand. Based on an extensive and consistent database and a standardized methodology, our approach will enable a high comparability of regional and global glacier dynamics and provide valuable information for the understanding of the role of mountain glaciers in respect to current and future sea level rise.

Area D - Board 483

Glaciers represent an important part of the hydrologic cycle in the Alps and they are very sensitive to climate change. Satellite remote sensing is an efficient tool for glacier monitoring because it provides a synoptic view of large areas. In literature, well-established methods for glacier delineation have been presented [1] [2], based on the Red / Short Wave Infrared (SWIR) ratio. These methods are dependent on a manual selection for each glacier of the best scene, i.e. absence of cloud coverage and minimum snow cover. A further manual refinement step is needed to handle possible errors in cloud cover or shadow area and debris covered ice.
A manual approach for glacier outline extraction, especially over large areas, may present relevant limitations for two main reasons [3]:
1) The increased amount of available satellite data provided by the recently launched Sentinel-2 mission, which ensure at least one acquisition every 5 days on a given area;
2) The need for a more frequent update of the glacier outlines i.e. few years, due to the faster changes affecting glaciers during the last years.
In this work we present a novel automatic method based on the full Sentinel-2 time series analysis for the delineation of bare ice and glacier accumulation zone.
The first step of the method is the classification of the Sentinel-2 images in five classes (clouds, snow, ice, water, other land cover) using a supervised Support Vector Machine classifier. Then, the best non-cloudy pixels with the lowest snow coverage in the surrounding area are selected by applying a moving window on the entire multitemporal image stack. The output is a multitemporal cloud free mosaic that represent the glaciers with the minimum snow cover. The last step of the method is the multitemporal mosaic classification into two classes, i.e. glacier, non-glacier, from which the glacier outlines can be easily extracted.
Coherence data from Sentinel-1 is then employed for debris covered glacier detection [4]: the surface flow of debris cover causes a lower coherence value with respect to other non-moving surfaces.
(Figure 1).
Figure 1- Block scheme of the proposed method for glacier outline extraction


The proposed approach for bare ice and accumulation zone outline detection from Sentinel-2 was firstly evaluated by visual interpretation over the Ortles-Cevedale group, the Adamello-Brenta Group and the Rofental (Figure 2). In this preliminary test, the proposed method can accurately detect the outlines of the glaciers in the investigated area.
Figure 2- Example of outlines of bare ice and accumulation area relative to the year 2017 over Hintereisferner.

A more rigorous validation procedure is planned within the project Interreg Italia-Österreich ITAT2025 GLISTT: An interregional glacier monitoring system for the Tyrol, South-Tyrol Region, where a team of glaciologists is providing reference outlines from orthophotos and very high resolution DEM acquired during different years. This reference dataset will also be employed to assess the performances of debris covered glacier detection by exploiting Sentinel-1 data.

References:
[1] F. Paul et al., “Guidelines for the compilation of glacier inventory parameters from digital sources,” Ann. Glaciol., vol. 50, no. January, pp. 119–126, 2010.
[2] K. J. Bayr, D. K. Hall, and W. M. Kovalick, “Observations on glaciers in the eastern austrian alps using satellite data,” Int. J. Remote Sens., vol. 15, no. 9, pp. 1733–1742, 1994.
[3] S. H. Winsvold, A. Kääb, C. Nuth, L. M. Andreassen, W. J. J. Van Pelt, and T. Schellenberger, “Regional Glacier Mapping Using Optical Satellite Data Time Series,” Cryosphere, vol. 12, no. 3, pp. 867–890, 2018.
[4] S. LIPPL, S. VIJAY, and M. BRAUN, “Automatic delineation of debris-covered glaciers using InSAR coherence derived from X-, C- and L-band radar data: a case study of Yazgyl Glacier,” J. Glaciol., vol. 64, no. 247, pp. 811–821, 2018.

Area D - Board 484

The surface flow of glaciers is a key component of a glacial system because it can be affected by changes in climatic conditions making it a direct indicator of glacier response to climate change. In a context of global climatic changes, it is then of general importance to accurately document variations in surface flow of glaciers worldwide. Another essential variable for the estimation of sea-level rise contribution and water resources is the thickness of the ice. Currently, the total ice volume stored in mountain glaciers is highly uncertain and estimates range from simple to double. Studies have shown that the accuracy of ice thickness estimations was strongly dependent on the quality of the input data and specifically the surface flow velocity (when used in the approach/model).
We present multi-temporal maps of glacier surface flow velocity for five mountain glaciers: Mittivakkat (Greenland), Mer de Glace (Alps), Khumbu (Himalaya), Artesonraju (Andes), Tasman glacier (Oceania) and the Astrolabe (Antarctica). These five target glaciers are located in different climate regions, hence covering a wide range of external forcings – atmospheric and/or oceanic – that may affect the changes in ice dynamics. These new data are used to (1) better document and understand the temporal variability of glacier flow at different time-scales (weekly to annual) and (2) improve glacier ice thickness estimates, an essential prerequisite for current and future glacier change impacts on water resources and sea level rise.
We use data from Sentinel-2 (ESA) and Landsat-8 (NASA-USGS) to derive ice-velocity on a weekly to annual basis and the high-resolution CNES’s Venμs (5-m) satellite with acquisition every two days and SPOT6/7 (1.5 m) and Pléiades (0.5 m) when available. We combine our ice velocity maps with existing ground observation datasets of surface elevations, surface mass balance (SMB) and GPS to understand the link between changes in ice dynamics, changes in SMB and basal sliding processes in the different climate regions. The data reveal the seasonal and annual variation of the glaciers with unequaled spatial and temporal resolution. Annual velocity maps are also generated by averaging the dense time series of surface flow obtained between the years 2013 and 2018 using Sentinel-2 and Landsat-8 data. These data are combined with local estimations of SMB, DEM’s and in-situ data to better constrain ice thickness distribution of selected glaciers using the method from Rabatel et al., 2018 and the mass conservation technique implemented in the Elmer/Ice model. The new ice thickness maps are evaluated using ground observations and compared to existing estimations from the Glacier Thickness Database in order to provide an accurate and distributed estimation of the glacier’s volume.

Area D - Board 486

Continental glaciers are witnesses of the changes undergone by our planet in the context of global warming. As in-situ data can sometimes be difficult to obtain, satellite data are key to observe and measure the changes on continental glaciers extent, height and volume. In particular, altimetry data can provide information on the glaciers surface height.
Long-term and qualitative altimetry data is an essential asset to study the world’s glaciers evolution, thanks to continuous measurements provided by the SRAL altimeters from ESA’s Copernicus Programme.

On current nadir altimeters, the Open-Loop Tracking Command (OLTC) allows to control the reception window of the return radar echoes. It consists in an on-board dedicated memory containing elevation information of the targeted surface for each on-orbit position. When the altimeter is operated in Open-Loop mode, its tracker is controlled by these tables and the reception window is set to record proper echoes of the reflective surface observed.
As it has been performed over inland waters, it is also key to optimize these on-board tables in mountainous regions where the tracker, if not properly controlled, does not allow to acquire the return signal from such specific surfaces that are glaciers. One of the challenges of continental ice observation is the glaciers surface slope, so this optimization must be performed carefully.

In this poster we show the method we used to define virtual stations over a dozen of glaciers overflown by Sentinel-3 A and Sentinel-3 B on their definitive orbits, in several continents: Europe (French Alps and Pyrenees), Asia (Pamir and Himalaya regions) and South America (Peru and Patagonia).
We illustrate with the very first results obtained over these particular examples and provide an overview of potential science results we can obtain from these altimetry measurements over continental glaciers.
This work will also present the technical opportunity of further optimizing on-board OLTC tables for users interested in obtaining altimetry data over glaciers.

Area D - Board 487

Rocky debris on glacier surfaces influences ice melt rates and the response of glaciers to climate change. However, scarce data on the extent and evolution of supraglacial debris cover have so far limited its inclusion in regional to global glacier models. Here we present an approach to automatically map supraglacial debris cover from optical satellite images at a global scale. Our approach makes use of the cloud-computing platform Google Earth Engine (GEE; https:// earthengine.google.com/) and exploits the large number of optical satellite images that are currently available.
In this contribution, we present mapping results from Landsat 8 and Sentinel-2 images, with 30 and 10 m spatial resolution, respectively. In principle, our approach allows rapidly mapping changes in the distribution and extent of debris cover at any time period, for which suitable satellite imagery is available in GEE, such as the Landsat data sets. The goal of this contribution is twofold. First, we present our new automatic mapping approach and evaluate it by comparison with a recently published data set of semiautomatically mapped supraglacial debris cover from the Karakoram and Pamir Mountains. Second, we provide the first global assessment of the spatial distribution of supraglacial debris cover, excluding the Greenland ice sheet and Antarctica, which are mostly free of debris.
We find that about 4.4% (~26,000 km2) of all glacier areas (excluding the Greenland ice sheet and Antarctica) are covered with debris, but that the distribution is heterogeneous. The largest debris-covered areas are located in high-mountain ranges, away from the poles. This likely reflects the fact that glaciers near the equator are exclusively found in high mountain ranges that are typically steep and feature abundant steep rock walls, that is, the source areas of the debris. In regions that are closer to the poles, ice cover increases and topographic elevations decrease to the extent that existing topography may even get buried by ice, hence eliminating any potential source areas for debris.
At a global scale, we find a negative scaling relationship between glacier size and percentage of debris. This observed global scaling of debris-cover fraction with glacier size supports the notion of source areas that get drowned with increasing glacier size. Therefore, the influence of debris cover on glacier mass balances is expected to increase in the future, as glaciers continue to shrink.

Area D - Board 488

Ocean tide induced modulation of ice flow rates of marine-terminating glaciers has been observed in Antarctica, Greenland and other glaciated regions. The largest ocean tides in Antarctica, with peak-to-peak amplitudes reaching nearly 6 meters, are located in the region of the Weddell Sea and strongly affect ice flow of the entire Filchner-Ronne Ice Shelf (FRIS) and its adjoining ice streams. Previous synoptic-scale observations of the tidal response of Rutford Ice Stream (RIS) using systematic Synthetic Aperture Radar (SAR) observations provided key insight into the mechanism of how ocean tidal signals are transmitted inland. Evans Ice Stream (EIS), also terminating in FRIS, is a significantly larger ice stream than RIS and encompasses many upstream tributaries. To infer the spatial and temporal variations in ice flow of EIS, we designed and initiated a dedicated observation plan of this region by COSMO-SkyMed (CSK) X-band SAR constellation. We also take advantage of the ongoing pre-defined mission observation scenario of Copernicus Sentinel-1ab (S1ab) at C-band. The tailored CSK acquisition campaign provides both ascending and descending HH-polarized StripMap HIMAGE scenes with high spatial resolution (3 m) and dense temporal sampling (with up to 7 observations/month in each viewing geometry). These data are complemented by S1ab, which provides synoptic views through descending mode Interferometric Wide swath (250 km) HH-polarized scenes with a 6 days revisit time. We first construct the secular 3D velocity field map, from which we derive strain-rate maps. We show the viability of combining the displacements derived from CSK and S1ab data with their different radar wavelengths, spatial resolutions and acquisition times. To evaluate our ability to resolve different families of tidal constituents, we develop tests using synthetic data consistent with the realized observation scenario. The joint use of CSK and S1ab data allows us to retrieve tidal signals with higher fidelity. Finally, we present the results of flow variations inferred from the actual data, including tidal response of both floating and grounded portions of ice.

Area D - Board 489