Changes in ice dynamics and mass balances on the northern derived from remote sensing data

Änderungen der Eisdynamiken und Gletschermassenbilanzen auf der nördlichen Antarktischen Halbinsel abgeleitet aus Fernerkundungsdaten

Der Naturwissenschaftliche Fakultät der Friedrich-Alexander-Universität Erlangen-Nürnberg zur Erlangung des Doktorgrades Dr. rer. nat. vorgelegt von Thorsten Christian Seehaus aus Bad Neustadt an der Saale Als Dissertation genehmigt von der Naturwissenschaftliche Fakultät der Friedrich-Alexander-Universität Erlangen-Nürnberg

Tag der mündlichen Prüfung: 29.11.2016

Vorsitzende/r des Promotionsorgans: Prof. Dr. Georg Kreimer

Gutachter/in: Prof. Dr. Matthias Braun

Prof. Dr. Wolfgang Rack

Prof. Dr. Helmut Rott Table of Contents 1 Summary...... 7 2 Zusammenfassung...... 9 3 Structure of the thesis...... 12 4 Remote sensing of polar ice sheets...... 13 4.1 SAR remote sensing...... 14 4.1.1 SAR Interferometry...... 17 4.1.1.1 Interferometric DEM processing...... 18 4.1.2 SAR offset tracking...... 19 4.1.2.1 Intensity tracking...... 19 4.1.2.2 Coherence tracking...... 19 4.2 Altimetry...... 20 4.3 Multispectral satellite remote...... 21 4.4 Ground penetrating radar sensing...... 21 4.5 Glacier mass balance derived from remote sensing data...... 21 4.5.1 Gravimetric mass balance method...... 22 4.5.2 Geodetic mass balance method...... 23 4.5.3 Input/Output mass balance method...... 23 5 Study site...... 24 5.1 Geography and climate ...... 25 5.2 Climatic and glaciological changes...... 26 6 Motivations and Objectives...... 29 6.1 Contributions to scientific papers...... 31 7 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier- system, Antarctic Peninsula...... 33 7.1 Introduction...... 34 7.2 Study site...... 36 7.3 Data...... 37 7.4 Methods...... 38 7.4.1 Glacier extent...... 38 7.4.2 Surface velocities...... 38 7.4.3 Surface elevations...... 38 7.4.4 Ice mass flux...... 39 7.4.5 Mass balance...... 39 7.5 Results...... 41 7.5.1 Glacier extent...... 41 7.5.2 Surface velocities...... 43 7.5.3 Surface elevation changes...... 43 7.5.4 Ice mass flux...... 44 7.5.5 Mass balance...... 45 7.6 Discussion...... 46 7.6.1 Velocity and elevation changes...... 46 7.6.2 Mass loss and Imbalance...... 47 7.7 Conclusions...... 49 8 Study 2: Dynamic response of Sjögren Inlet , Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series...... 51 8.1 Introduction...... 52 8.2 Study site...... 53 8.3 Data and Methods...... 54 8.3.1 Flow velocity and glacier extent...... 54 8.3.2 Surface elevation changes...... 54 8.3.3 Mass balance...... 56 8.3.4 Ice mass flux...... 57 8.4 Results...... 57 8.4.1 Flow velocity and glacier extent...... 57 8.4.2 Surface elevation changes...... 59 8.4.3 Mass balance...... 61 8.4.4 Ice mass flux...... 62 8.5 Discussion...... 63 8.5.1 Ice dynamics...... 63 8.5.2 Mass balance...... 66 8.6 Conclusions...... 67 9 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985...... 70 9.1 Introduction...... 71 9.2 Study site...... 72 9.3 Data & Methods...... 73 9.3.1 Area Changes...... 73 9.3.2 Flow velocity...... 74 9.3.3 Ice mass flux...... 76 9.3.4 Glacier attributes...... 77 9.3.5 Cluster analysis...... 77 9.3.6 Climatic mass balance...... 77 9.4 Results...... 78 9.4.1 Area changes...... 78 9.4.2 Flow velocity...... 80 9.4.3 Ice mass flux...... 82 9.4.4 Glacier attributes...... 83 9.4.5 Cluster analysis...... 83 9.4.6 Climatic mass balance...... 84 9.5 Discussion...... 84 9.5.1 Ice dynamics...... 84 9.5.1.1 East-Ice-Shelf...... 85 9.5.1.2 East...... 86 9.5.1.3 West...... 87 9.5.2 Mass balance...... 89 9.6 Conclusions...... 91 10 Conclusions and Outlook...... 93 11 Acknowledgements...... 95 12 References...... 96 13 Appendix...... 107 13.1 Supplemental material to Study 1 ...... 107 13.2 Supplemental material to Study 3...... 122 Summary

1 Summary

The climate system of our planet has undergone significant changes since the beginning of industrialization and the global average surface temperature has risen by 0.74±0.18°C per century. The cryosphere plays an important role in the global climate system. It shows considerable changes in response to climatic variations. Large amounts of water are stored in the glaciers and ice sheets. However, the changes in climatic conditions lead to shrinkage of the ice masses, resulting in a contribution to the global sea level rise. The Antarctic Ice Sheet has shown significant changes in the last century and major ice mass loss was reported in West and along the Antarctic Peninsula. The Antarctic Peninsula is a hotspot of global warming and is affected by dramatic climatic and glaciological changes. A strong temperature increase was observed in the 20th century. Along the peninsula numerous ice shelves have retreated considerably or even disintegrated within the last decades. The recession or break-up of the ice shelves has led to a reduced buttressing of the tributary glaciers. Consequently, the ice flow of former tributaries has accelerated, resulting in increased mass discharge. Widespread recession of tidewater glaciers was also reported along the Antarctic Peninsula, and surface lowering and speed-up of the outlet glaciers was found on the west coast. Since the 1960s, an increasing amount of spaceborne remote sensing data has been acquired worldwide. Field observations in polar regions are expensive, laborious and usually temporally and spatially limited. Thus, remote sensing data facilitates the spatially extended investigation of the large polar ice sheets, especially in inaccessible areas, complementary to field campaigns. Numerous methods have been developed in order to analyze the changes of glaciers and ice sheets based on remote sensing data. The Antarctic Peninsula has difficult terrain and is frequently covered by clouds. Hence, radar remote sensing is an ideal tool to investigate this region, since the radar signal penetrates through clouds. In 1991, the European Remote Sensing satellite mission was launched by the European Space Agency, and has frequently acquired radar remote sensing data of the polar regions. This was followed by more radar satellite missions. On the north eastern Antarctic Peninsula, the Larsen-A and Prince-Gustav ice shelves disintegrated in the 1990s. Thus, the available radar remote sensing time series allows us to observe these events and the resulting long-term impacts on the tributary glaciers in detail. Two comprehensive studies were performed in order to investigate the glaciers' response to ice shelf decay; the first was conducted at the Dinsmoor-Bombardier-Edgeworth (DBE) glacier system, former tributaries of the Larsen-A Ice Shelf and the second was conducted at the Sjögren Inlet glaciers, former tributaries of the Prince-Gustav Ice Shelf. Detailed time series of glacier flow speed, surface elevation change, frontal retreat and ice discharge were derived from various satellite datasets. Moreover, glacier mass balances were determined for different time periods and by applying different methods and assumptions. The tributaries showed a strong acceleration of ice flow (up to 8.8 m/d at DBE) and fast ice front recession after the ice shelf break-up, followed by a slowdown and stabilization or re-advance of the calving front. An exponential surface lowering

7 Summary trend was obtained for the DBE glacier system, amounting to 130±15 m in the time period 1995- 2014, whereas the Sjögren Inlet glaciers revealed nearly linear trends of 68±10 m on average. Highly negative mass balances were found in the first years after ice shelf collapse (DBE: -4.86±0.49 Gt/a in 1995-2003; Sjögren Inlet: −1.16 ± 0.38 Gt/a in 1993-2001), which decreased towards recent years (DBE: -0.17±0.15 Gt/a in 2003-2014; Sjögren Inlet: −0.54 ± 0.13 Gt/a in 2012-2014). However, the glaciers still showed increased flow speeds as well as considerable mass loss and surface lowering rates in 2014, about 20 years after ice shelf removal. Variations in the timing and intensity of the glaciers response to similar forcing were attributed to differences in the fjord and glacier geometries. Numerous ice shelves around Antarctica are thinning and retreating. Thus, projections of the reaction of tributaries to potential ice shelf break-up need detailed information about the ongoing processes. Therefore, the obtained results provide a consistent long-term database of the dynamic adjustments of tributary glaciers to ice shelf disintegration. In a third study, glacier changes since 1985 along the west and east side of the northern Antarctic Peninsula were investigated. Temporally detailed time series analysis of ice velocities and ice front positions were performed at 74 glacier basins north of 65°S. The information obtained on changes in ice dynamics were discussed for sub-regions and linked with atmospheric, oceanic and glaciological conditions. For instance, glaciers affected by ice shelf disintegration showed similar variations and the termini of all basins along the east coast retreated. On the west side, the temporal trends did not indicate a clear change pattern. However, the observed variations in ice dynamics could be correlated with the geometric parameters of the individual basins by applying a cluster analysis. Moreover, glacier mass balance of the study site was computed for two epochs. On the east coast negative mass balances in the 1990s and a nearly balanced state in the period 2010-2014 were found. In contrast, very high accumulation rates were obtained along the west side, resulting in positive mass balances. Hence, the recent mass balance of the northern Antarctic Peninsula is estimated to be close to equilibrium or even positive, depending on the applied ice thickness and climatic mass balance scenario. Area, flow speed and surface elevation changes can be determined at good quality from remote sensing data. However, the performed studies indicate that the climatic mass balance and ice thickness information are the major sources of uncertainty in imbalance computations on the Antarctic Peninsula. Observations of both are very sparse on the Antarctic Peninsula due to the harsh conditions found there. Computer models already provide good and region-wide information, but their spatial resolution and accuracies are limited. Therefore, more observations and knowledge of the ongoing processes are necessary in order to validate and further improve the current models and mass balance estimates.

8 Zusammenfassung

2 Zusammenfassung

Das Klima auf unserer Erde hat sich seit Beginn der Industrialisierung signifikant verändert und die durchschnittliche globale Oberflächentemperatur ist um 0,74 ± 0,18 ° C pro Jahrhundert gestiegen. Die Kryosphäre spielt eine wichtige Rolle im globalen Klimasystem und zeigt deutliche Reaktionen auf den Klimawandel. Große Mengen an Wasser sind in den Gletschern und Eisschilden gespeichert. Allerdings führen die klimatischen Veränderungen zu einem Rückgang der Eismassen, und tragen somit zum globalen Meeresspiegelanstieg bei. Innerhalb des letzten Jahrhunderts zeigte das antarktische Eisschild deutliche Massenänderungen, wobei die größten Eismassenverluste in der Westantarktis und entlang der Antarktischen Halbinsel beobachtet wurden. Die Antarktische Halbinsel ist ein Brennpunkt der globalen Erwärmung mit dramatischen klimatischen und glaziologischen Veränderungen. Im 20. Jahrhundert wurde ein weltweit überdurchschnittlicher Temperaturanstieg beobachtet und zahlreiche Schelfeise entlang der Antarktischen Halbinsel zogen sich stark zurück oder lösten sich in den letzten Jahrzehnten ganz auf. Der Rückgang und Aufbruch der Schelfeise führte zu einer reduzierten Abstützung der Zuflussgletscher. In Folge dessen beschleunigten die ehemaligen Zuflussgletscher, wodurch es zu erhöhtem Eismassenabfluß kam. Auch die direkt ins Meer kalbenden Gletscher zogen sich im vergangenen Jahrhundert entlang der Antarktischen Halbinsel zurück. An der Westküste kam es zu Höhenabnahmen und Beschleunigungen der Gletscher. Seit den 1960er Jahren wird eine große Menge verschiedenster Satelliten-Fernerkundungsdaten aufgezeichnet. Feldbeobachtungen in den Polarregionen sind teuer, aufwendig und in der Regel zeitlich und räumlich begrenzt. Folglich erleichtert die Analyse von Fernerkundungsdaten die räumlich ausgedehnte Untersuchung der polaren Eismassen, vor allem in schwer zugänglichen Gebieten, und bietet eine ideale Ergänzung zu Feldmessungen. Zahlreiche Methoden wurden entwickelt um die Veränderungen der Gletscher und Eisschilde mittels Fernerkundungsdaten zu analysieren. Die Antarktische Halbinsel ist in weiten Teilen nur sehr schwer zugänglich und zudem häufig von Wolken bedeckt. Daher ist die Radar-Fernerkundung ein ideales Werkzeug um diese Region zu untersuchen, da das Radarsignal die Wolkendecke durchdringt. Im Jahr 1991 wurde die European Remote Sensing Satellitenmission der Europäischen Weltraumagentur gestartet, welche eine Vielzahl an Radarfernerkundungsdaten der Polarregionen lieferte. Mehrere Radarsatellitenmissionen folgten. Im Nordosten der Antarktischen Halbinsel zerfielen das Larsen-A und Prinz-Gustav Schelfeis in den 1990er Jahren. Somit bieten die verfügbaren Radarfernerkundungsdaten eine sehr gute Datenbasis um diese Schelfeisaufbrüche und die daraus resultierenden langfristigen Auswirkungen auf die Zuflussgletscher zu analysieren. Zwei umfassende Studien wurden durchgeführt um die Reaktionen der Gletscher auf Schelfeisaufbrüche zu untersuchen. Die erste Studie wurde am Dinsmoor-Bombardier-Edgeworth (DBE) Gletschersystem, ehemalige Zuflüsse des Larsen-A Schelfeises, durchgeführt und die zweite Studie an den Sjögren Inlet Gletschern, ehemalige Zuflüsse des Prinz-Gustav Schelfeises. Detaillierte Zeitreihen von Gletscherfließgeschwindigkeiten, Oberflächenhöhenänderungen,

9 Zusammenfassung

Rückzug der Eisfront und des Eismassenabflusses wurden aus verschiedenen Satellitendatensätzen abgeleitet. Darüber hinaus sind die Gletschermassenbilanzen für unterschiedliche Zeiträume und unter der Verwendung verschiedener Methoden und Randbedingungen bestimmt worden. Die Zuflussgletscher zeigten eine starke Beschleunigung (auf bis zu 8,8m/d am DBE) und einen rapiden Rückzug der Gletscherzungen als unmittelbare Folgen des Schelfeisaufbruchs. Anschließend kam es jedoch zu einem Rückgang der Fließgeschwindigkeiten und Stabilisierung oder Wiedervorstoßen der Gletscherzuge. Ein exponentieller zeitlicher Verlauf der Höhenänderung wurde am DBE Gletschersystem beobachtet, mit einer Gesamthöhenabnahme von 130±15m im Zeitraum 1995-2014. Die Gletscher im Sjögren Inlet hingegen zeigten fast lineare Höhenänderungstrends von durchschnittlich -68±10m zwischen 1993 und 2014. In den ersten Jahren nach den Aufbrüchen der Schelfeise wurden stark negative Massenbilanzen beobachtet (DBE: -4,86±0,49Gt/a im Zeitraum 1995-2003; Sjögren Inlet: -1,16±0,38Gt/a im Zeitraum 1993-2001), welche in den letzten Jahren deutlich zurückgegangen sind (DBE: -0,17±0,15Gt/a im Zeitraum 2003-2014; Sjögren Inlet: -0,54±0,13Gt/a im Zeitraum 2012-2014). Jedoch zeigten die Gletscher im Jahr 2014, etwa 20 Jahre nach den Schelfeisaufbrüchen, immer noch erhöhte Fließgeschwindigkeiten, sowie erhebliche Massenverluste und Höhenabnahmen. Die Unterschiede im zeitlichen Ablauf und in der Intensität der Reaktionen der Gletscher auf die Schelfeisaufbrüche sind Unterschieden in der Fjord- und Gletschergeometrien zugeordnet worden. Zahlreiche Schelfeise in der Antarktis ziehen sich zurück und verlieren an Eisdicke. Für Prognosen von Zuflussgletscherreaktionen auf potenzielle Schelfeisaufbrüche werden detaillierte Informationen der ablaufenden Prozesse benötigt. Somit bieten die gewonnenen Ergebnisse eine konsistente Datenbank mit Informationen über die dynamischen, langfristigen Reaktionen von Zuflussgletschern auf Schelfeisaufbrüche. In einer dritten Studie wurden die Gletscheränderungen seit 1985 entlang der West- und Ostseite der nördlichen Antarktischen Halbinsel untersucht. Detaillierte Zeitreihenanalysen von Fließgeschwindigkeiten und Kalbungsfront-Positionen wurden nördlich von 65°S an 74 Gletschern durchgeführt. Die gewonnenen Informationen über Veränderungen in den Eisdynamiken sind mit atmosphärischen, ozeanischen und glaziologischen Variationen korreliert und für einzelne Subregionen diskutiert worden. Die vom Zerfall der Schelfeise betroffenen Gletscher zeigten weitgehend ähnliche Veränderungen und alle Gletscher entlang der Ostküste zogen sich zurück. Auf der Westseite zeigten die zeitlichen Trends der Eisdynamiken keine eindeutigen Muster. Jedoch konnten die beobachteten Änderungen der Eisdynamiken mit geometrischen Parametern der einzelnen Gletscherbecken durch Anwenden einer Clusteranalyse korreliert werden. Weiterhin wurde die Gletschermassenbilanz des Untersuchungsgebietes für zwei Zeitabschnitte berechnet. An der Ostküste sind deutlich negative Massenbilanzen in den 1990er Jahren und nahezu ausgeglichene Massenbilanzen in der Periode 2010-2014 gefunden worden. Im Gegensatz hierzu wurden entlang der Westseite sehr hohe Akkumulationsraten beobachtet, welche zu positiven Massenbilanzen führten. Somit ergibt sich eine ausgeglichene oder sogar positive Massenbilanz für die nördliche Antarktische Halbinsel im Zeitraum 2010-2014, je nach verwendetem Eisdicken- und Akkumulationsszenario. Änderungen der Gletscherausdehnung, Fließgeschwindigkeit und Oberflächenhöhe können mittels Fernerkundungsdaten mit guter Qualität bestimmt werden. Allerdings heben die durchgeführten

10 Zusammenfassung

Studien hervor, dass die klimatischen Massenbilanz- und Eisdickeninformationen die größten Fehlerquellen bei Bilanzierungen der Eismassenänderungen auf der Antarktischen Halbinsel sind. Messungen beider Variablen sind aufgrund der harschen Bedingungen auf der Antarktischen Halbinsel sehr spärlich. Computermodelle liefern zwar bereits gute und flächendeckende Datensätze, aber ihre räumliche Auflösung und Genauigkeit sind begrenzt. Daher sind weitere Beobachtungen und Erkenntnisse über die ablaufenden Prozesse notwendig, um die aktuellen Modelle und Massenbilanzschätzungen zu bestätigen und weiter zu verbessern.

11 Structure of the thesis

3 Structure of the thesis

The aim of this thesis was to study the changes in ice dynamics and glacier mass balances on the northern Antarctic Peninsula in the last decades, by analyzing data from various remote sensing missions. Firstly an overview of remote sensing of polar ice sheets is provided in Section 4. The history of technological evolution of air and spaceborne remote sensing sensors, used to study the polar regions, is briefly described. Afterwards, the sensors and methods applied in this work are introduced. The analyses in this work are mainly based on Synthetic Aperture Radar remote sensing. Therefore, the general principle of this method and the applied data processing techniques are explained in detail. In Section 5, an introduction to the study site, the Antarctic Peninsula, is provided. In order to give an overview on the settings of the study region, the geography and the general glaciological and climatic conditions on the Antarctic Peninsula are explained. Subsequently, the variations in climatic and glaciological conditions are presented, demonstrating the observed dramatic changes and the scientific state of the art at the study site. Based on this, the motivations and objectives of this work are introduced. Three studies were conducted in order to investigate the target region concerning the defined objectives. The outcome of the surveys had been summarized in scientific papers and short summaries and the author's contribution to the individual publications is listed in Section 6.1. In the following sections the performed studies are explained in detailed by means of the related scientific manuscripts. The first two studies were focused on detailed analyses of the reaction of tributary glaciers to ice shelf disintegration on the northeastern Antarctic Peninsula. The glaciological changes of 74 glacier basins along the northern Antarctic Peninsula were determined in a third study, in order to investigate the ongoing processes of a wider region as well as to take into account the changes on the west side of the Antarctic Peninsula. Finally, a comprehensive conclusion of the results of the presented studies and an outlook is presented in Section 10.

12 Remote sensing of polar ice sheets

4 Remote sensing of polar ice sheets

The polar ice sheets play a significant role in the global climate system and are strongly influenced by climate change. Due to their remote location and harsh climate conditions, in-situ measurements are very labor-, time- and cost-intensive and even partly dangerous. The polar ice sheets extend over millions of square kilometers, but in-situ observations mostly just cover relatively small areas. Therefore, remote sensing applications are well appropriate to study these regions with higher frequency and larger spatial extent, complementary to in-situ campaigns. Remote sensing of the polar regions started in the 1920s. National mapping agencies focused on strategic topographic mapping of the widely unknown areas by aerial photography (Pope et al., 2014). The era of space-borne remote sensing began shortly after the launch of the first satellite, “Sputnik 1” in October 1957. Many satellite photography campaigns were carried out during the Cold War. Declassified images from various surveillance missions, such as the Argon or Corona Programs, were made available afterwards. By using modern image analysis techniques, the historical archives of aerial and satellite photography can provide important information on the ancient state of the ice sheets (e.g. glacier extents and surface structure) (Pope et al., 2014). In 1972 the well-known Landsat Program was started by the National Aeronautics and Space Administration (NASA) in the US. Since more than 40 years, it provides multispectral Earth observation data, which is widely-used by scientists from different disciplines. Various multispectral sensor missions followed, e.g. AVHRR in 1978, MODIS & ASTER in 1999, Sentinel-2 in 2016. The body of knowledge obtained from multispectral imagery in cryospheric science is enormous. A huge amount of different applications and techniques have been developed to study changes of the ice sheets, such as monitoring of ice flow, mapping of glacier extents, and computation of surface elevation changes (e.g. Berthier et al., 2012; Burton-Johnson et al., 2016; Haug et al., 2010; Rankl et al., 2014). The first space-borne Synthetic Aperture Radar (SAR) sensor was aboard the Seasat satellite in 1978. Though the satellite operated only for about 100 days, the capability of orbital SAR instruments to obtain valuable Earth observation data was proved. The European Space Agency (ESA) launched the very successful satellite programs ERS-1 in 1991, ERS-2 in 1995 and Envisat in 2002 (Rees, 2006). Large amounts of SAR data were acquired worldwide and some specific mission phases were performed in order to study the cryosphere, e.g. the first and second Ice phase in 1991/1992 and 1993/1994, respectively and the ERS-1&2 Tandem phase in 1995/96. Numerous SAR missions from other space agencies followed after the start of ERS-1, e.g. RADARSAT-1 in 1995, ALOS PALSAR in 2006 and TerraSAR-X in 2007 and different radar frequencies were applied. The SAR sensors were continuously improved and the number of imaging and polarization modes as well as spatial resolution increased (Pope et al., 2014). Various techniques were developed to study snow and ice. Very important SAR techniques used to study glacier mass balances are offset tracking and interferometry. The key parameters of ice sheet changes (surface elevation change, ice velocity, grounding line location, calving front location) can be obtained from adequate SAR acquisitions using these methods (Polar Space Task Group, 2016). Therefore, the importance of satellite SAR data to study the ice sheets has more and more increased in the last two decades. Moreover, SAR acquisitions of the polar regions were efficiently 13 Remote sensing of polar ice sheets coordinated during the International Polar Year 2007/08 by the Space Task Group, and many outstanding insights on the ice sheets were obtained, e.g. ice velocity maps of (central) Antarctica and Antarctic grounding line from differential satellite radar interferometry (Rignot et al., 2011a, 2011b; Scheuchl et al., 2012). Based on this successful cooperation of various international space agencies and scientific unions, the Polar Space Task Group (PSTG) was established in 2011 as an Advisory Group to the World Meteorological Organization Executive Council Panel of Experts on Polar Observations Research and Services (EC-PORS). In regards to SAR acquisitions, the SAR Coordination Working Group, a PSTG sub-group, focuses on the specific requirements and coordinates data acquisitions of the polar ice sheets. Their aim is to enhance the knowledge of these remote regions by using SAR data (Polar Space Task Group, 2016). In addition to imaging sensors, air and spaceborne altimetry provides an important source of information on the state of the ice sheets. Nadir looking radar altimeters were part of the ERS-1/2 and Envisat satellite missions (Rees, 2006). CryoSat-2 has been in orbit since 2010, having followed the failed launch of CryoSat-1 in 2005. Its primary instrument is the Synthetic Aperture Interferometric Radar Altimeter-2 (SIRAL-2). It is very suitable for observing surface elevation changes of marine ice sheets and sea ice thickness and provides altimeter measurements up to latitude of 88° S/N with very high across-track spacing compared to previous sensors (Helm et al., 2014; Wingham et al., 2006). The Geo science Laser Altimeter System (GLAS) aboard the Ice, Cloud and land Elevation Satellite (ICESat) acquired data from 2003 to 2009, which is widely used to analyze ice surface elevation changes (e.g. Bolch et al., 2013; Pritchard et al., 2009; Zwally and Giovinetto, 2011). Its successor, ICESat-2, was planned to be launched in 2015 (postponed to 2017) and will provide multi-track laser altimetry data (Abdalati et al., 2010). Remote sensing data types and methods applied in this work to observe glaciological changes are discussed in more detail in the following sections. Table 4-1 provides an overview of the air and spaceborne sensors used. 4.1 SAR remote sensing The history of RADAR (RAdio Detection And Ranging) dates back to the late 19th century. It is an active microwave system consisting of a transmitter and a receiver. A pulse of electromagnetic energy is sent towards the target and based on the echo delay time, its distance from the Radar system can be calculated. The first Radar instruments were applied for naval and military purposes. After World War II, civil and scientific applications of Radar systems were initiated (Watson, 2009). Typically, a radar system for remote sensing tasks is mounted on board an air or spacecraft. It transmits microwave pulses of electromagnetic energy at specific wavelengths perpendicular to the flight direction (side-looking radar) towards the Earth's surface and detects the return signals and their traveling time (Figure 4-1). The most commonly used wavelengths (e.g. K, X, C, L) were named by the military during World War II ranging between 1 cm and 100 cm (see also Table 4-1). The resolution in look direction (range direction) is proportional to the microwave pulse length. Short pulses lead to higher range resolutions but the energy to detect the objects and consequently the reflected energy is also reduced. Thus, a balance between higher resolution and transmitted energy must be found. However, by using radar pulses with a frequency shift over time, this technique is called “chirp”, the range resolution can be increased without reducing the pulse length (Woodhouse, 2006). In the along-track direction (azimuth direction), the resolution depends 14 Remote sensing of polar ice sheets on the beam width of the transmitted signal and is inversely proportional to the antenna length. Higher azimuth resolutions can be obtained by using longer antennas (Jensen, 2007). As space on board of an airplane or satellite is limited, Carl A. Wiley invented the Synthetic Aperture Radar in 1951, which was further developed in the following years (Love, 1985). The SAR sensor on board an aircraft or satellite moves relative to the observed object, and receives its radar echoes in multiple antenna positions. The successive recordings are combined by applying a sophisticated signal processing technique and a large synthetic aperture (antenna length) is formed. This signal processing leads to a finer imaging resolution as compared to possible resolutions based on the physical parameters of the antenna. In contrast to optical sensors, SAR systems are independent of solar light and the radar signal penetrates through clouds. Therefore, SAR remote sensing is extremely valuable in polar regions with long lasting polar nights and areas with frequent cloud covers like the Antarctic Peninsula.

S1

S2

Figure 4-1: Geometry of an InSAR configuration. The SAR sensors (S1 & S2) are separated by the baseline B. R1/R2: range between the SAR sensors and the observed object; Bp: perpendicular baseline; α: Angle between horizontal direction and baseline B; θ: incidence angle; dh: elevation difference between neighboring objects

However, due to the side-looking geometry, SAR sensors have also some limitations. Geometric distortions in SAR imagery are caused by the topography of the mapped region. The inclination of the surface affects the local incidence angle of the radar signal (angle between a vertical line on the ground and a line from the sensor to the object on the ground). A slope towards the radar (foreslope) leads to a shortening and an inclination away from the sensor (backslope) to an elongation of the real feature distances (ground-range) in the SAR acquisition geometry (slant- range). If the incidence angle is smaller than the foreslope, the top of the mountain reflects the microwave back to the sensor before the lower parts of the slope are hit by the SAR signal. Hence, the mountain slope appears to overhang in slate-range geometry, which is called a “layover”. Additionally, if the backslope is steeper than the depression angle (90° minus incidence angle), it is

15 Remote sensing of polar ice sheets in radar shadow. Both layovers and shadowing lead to a loss of information and cannot be corrected. Radar systems are a coherent electromagnetic radiation source. The random interference of the microwaves produces random bright and dark spots in SAR images, also called “speckles”. Multilooking is a common procedure to reduce speckles. Adjacent pixels (looks) in a defined window are averaged, though, leading to a lower image resolution (Jensen, 2007; Massom and Lubin, 2006). However, speckles can also be useful in order to obtain information on surface displacements between consecutive SAR acquisitions (“speckle tracking” see Section 4.1.2.1). Table 4-1: Overview of sensors used in this work. a) SAR sensors TerraSAR-X Mission ERS-1/2 Envisat Radarsat-1 ALOS TanDEM-X Sensor SAR ASAR SAR PALSAR SAR Agency ESA ESA CSA JAXA DLR Revisit time [d] 35 35 24 46 11 Wavelength [cm] 5.6 5.6 5.6 23.5 3.1 Radar Band C C C L X Launch 1991/1995 2002 1995 2006 2007/2010 Mission End 2000/2011 2012 2013 2011 -

b) Multispectral sensors Mission Landsat 5 Terra SPOT 5 Sensor TM ASTER HRS Agency NASA NASA CNES Revisit time [d] 16 16 26 Resolution [m] 30/120 15/30/90 2.5/5/10/20 Launch 1984 1999 2002 Mission End 2013 - -

c) Altimeter Mission ICESat CryoSat-2 Operation IceBridge Sensor GLAS SIRAL ATM Agency NASA ESA NASA Type Laser (space-borne) SAR (space-borne) Laser (airborne) Wavelength 532/1064 nm 2.2 cm 532 nm Footprint [m] 70 200 1 Launch 2013 2010 2009 Mission End 2010 - -

16 Remote sensing of polar ice sheets 4.1.1 SAR Interferometry SAR data does not just contain information about the intensity of the reflected signal. The phase information is also inherent due to the coherent nature of the radar signal and the coherent signal processing. Caused by different signal traveling times, phase differences between SAR acquisitions, mapped with slightly offset antenna positions, can be correlated. This technique is called interferometric SAR (InSAR) and its principle is illustrated in Figure 4-1. An interferogram is computed by multiplying the complex signal of the first SAR image with the complex conjugate of the later image. The scattering mechanism of the target on the surface is assumed to be the same, resulting in a phase difference dФ which is proportional to the slant-range differences dR between both data sets.

d Φ=2π/λ dR (1)

λ: wavelength of SAR signal Two basic InSAR concepts exist: repeat pass interferometry and single-pass interferometry (Rosen et al., 2000; Zebker and Goldstein, 1986). The first concept, repeat pass interferometry, uses repeated acquisitions of the same area from slightly different antenna positions. The orbits of a satellite are usually displaced by several hundred meters in each repeat cycle in the across-track direction. The shift of the satellite orbits (determined from satellite orbit data) and also displacement (in range direction) of the scattering phase center on Earth can lead to slant-range differences dR. Therefore subsequent images (separated in time) of the same region and satellite orbit can be interferometrically processed to analyze the Earth's surface. Various specific repeat pass interferometry techniques (e.g. 2, 3 or 4 path differential InSAR) were developed to study processes such as ice dynamics, land deformation and land cover changes (Fletcher and European Space Agency, 2007). In glaciology, typical applications are grounding line mapping, ice motion detection and surface elevation change measurements (e.g. Rack and Rott, 2004; Rankl and Braun, 2016; Rignot et al., 2011b). However, the time difference between the acquisitions (temporal baseline) also limits the applicability of repeat pass InSAR, especially for cryospheric purposes. Significant changes of the surface such as snow fall or melt as well as high ice flow velocities cause changes in the backscattered signal, leading to a loss of coherence. Consequently, no information of the phase differences can be obtained (Massom and Lubin, 2006). On the other hand, this can be also useful, e.g. for mapping glacial extent (Atwood et al., 2010). For single-pass interferometry the reflected SAR signal is simultaneously recorded by two separated antennas in the across-track direction (perpendicular to the flight direction). Consequently, the slant-range difference dR is only caused by the different positions of the antennas (Pellikka and Rees, 2010). Single-pass InSAR is suitable for the generation of digital elevation models (DEMs) and is therefore very useful for observing elevation changes in the polar regions. The first space-borne across-track InSAR mission was the Shuttle Radar Topography Mission (SRTM) in 2000, lasting 11 days. It had a fixed baseline B of 60 m between both antennas and just covered regions between 60°N and 54°S. The polar regions were not mapped by SRTM. The TanDEM-X mission has run since 2010 and provided high-resolution single-pass InSAR data with variable baseline length B. The TanDEM-X data is highly valuable for studying the polar ice 17 Remote sensing of polar ice sheets sheets. Changes in ice surface elevation were derived from TanDEM-X data in this work, and the generation of DEMs with InSAR data is described in the next section (Fletcher and European Space Agency, 2007; Fritz et al., 2011). 4.1.1.1 Interferometric DEM processing As discussed above, the temporal decorrelation due to surface changes (movements or variations of the scattering properties) is minimized using single-pass InSAR data and thus, it is ideal to generate DEMs of glaciated regions and to monitor their surface elevation changes by repeated coverage. Additionally, phase differences due to variations of atmospheric conditions as well as phase noise are also negligible. Considering the geometry in Figure 4-1 and applying the parallel- ray approximation (B<

dR=B sin (θ−α) (2)

α: Angle between horizontal direction and baseline B θ: incidence angle Thus, the phase difference dФ (Eq. 1) can be expressed for the bistatic case as: 2π 2 π d Φ= B sin (θ−α)= B λ λ P (3)

Bp: perpendicular baseline and the relative phase difference ΔdФ between two locations caused by their different topographic heights dh results in (Rosen et al., 2000):

2π BP dh Δ d Φ= λ (4) R 1 sinθ

This relation is also called phase-to-height sensitivity. The relative phase difference between different locations is not unambiguous and can be biased by a multiple of 2π since dΦ is wrapped in the interval 0-2π. Therefore, the relative phase differences between neighboring pixels in the SAR data need to be summed up to obtain the absolute phase difference relative to a reference point with known altitude. This process is called phase unwrapping (Rosen et al., 2000). If the relative phase difference due to the elevation difference between two adjacent pixels is more than 2π, phase jumps occur in the unwrapped interferogram causing errors in the final DEM product. Different phase unwrapping algorithms can be used in order to avoid or minimize the appearance of phase jumps. Finally, the unwrapped absolute phase difference is transferred into terrain heights using the phase-to-height sensitivity (Equation 4). The TanDEM-X mission consists of two nearly identical X-Band SAR satellites forming a spaceborne interferometer. The first, TerraSAR-X, was launched in June 2007. Its companion, TanDEM-X, followed in June 2010. Since that time, both satellites fly in a close helix-like formation acquiring single-pass InSAR data (Fritz et al., 2011). The processing of TanDEM-X data is performed in a slightly different way. A differential interferogram is calculated by means of a reference DEMs and the InSAR data. This differential approach is applied in order to further suppress errors in the final DEMs caused by phase jumps. For more details of the processing chain, see Sections 7.4.3 and 13.1.

18 Remote sensing of polar ice sheets 4.1.2 SAR offset tracking The motion of glacier surfaces can be studied by analyzing sequential SAR data. Interferometric SAR procedures to observe glacier flow have significant shortcomings, such as temporal decorrelation or displacement measurement only in the slant range direction (Pellikka and Rees, 2010). Therefore, SAR offset tracking techniques are excellent tools to monitor ice surface velocities, especially intensity tracking for fast flowing glaciers. The procedures need as input data two consecutive coregistered SAR datasets with similar acquisition parameters, e.g. radar wavelength, satellite path, incident angle. 4.1.2.1 Intensity tracking Features that are visible in the intensity images, such as crevasses or surface structures, are tracked by this technique to obtain ice surface velocities. The SAR signal penetrates into snow and ice. Thus, more features, such as internal inhomogeneities and crevasses, can be detected as compared to optical imaginary (Massom and Lubin, 2006). The tracking is done by cross- correlating the backscatter intensity of patches (chips or windows) in a SAR image with patches in the consecutive acquisition. A 2-dimensional cross-correlation field between the reference window and the search area in the second image is computed in order to obtain the displacement of the image section. The search area is larger than the window size. Numerous patches are regularly defined in the search area and cross-correlated with the reference patch (moving window technique). The cross-correlation is performed by a conjugate multiplication of Fast Fourier Transforms (FFT) of the image patches, since the computation in spatial frequency space is more computationally efficient. The offset between the center of the reference window and the maximum in the cross-correlation field refers to the displacement of the observed area between both image acquisitions. In order to increase the accuracy, a 2-dimensional regression function is fitted to the cross-correlation field, and its peak is used for the offset estimation. Typically, this procedure is performed on numerous regular gridded locations across the SAR acquisitions to obtain a displacement map. The displacement is measured in slant-range geometry and needs to be transferred in ground-range geometry to obtain the real surface velocity field. The signal-to-noise ratio (SNR) between the modeled peak and the mean of the surrounding correlation field is an indicator of the quality and reliability of the offset measurement. Offset estimates with SNR below a defined threshold are rejected and only good quality measurements remain (e.g. de Lange et al., 2007; Scambos et al., 1992; Strozzi et al., 2002). Often, additional filtering techniques are applied in order to further increase the quality of the results (e.g. Burgess et al., 2012; Pritchard, 2005). In areas with few surface features and/or very low flow velocities, the speckle pattern can be used to track the displacement if coherence is retained between both acquisitions. The principles are the same as for the intensity tracking, however, smaller tracking window sizes are needed. The accuracies achieved are up to an order of magnitude higher as compared to the tracking of visible features (Gray et al., 1998). 4.1.2.2 Coherence tracking When the displacement and the changes in the surface structure are small, coherence is retained between the SAR acquisitions and surface motion can be determined from the phase information of the complex SAR data (Pellikka and Rees, 2010). As for the intensity tracking method, a moving

19 Remote sensing of polar ice sheets window algorithm is applied to infer the displacement at certain locations in the SAR images. However, the coherence between the patches is computed by generating small interferograms instead of the cross-correlating the intensity. The peak of the coherence field is obtained by means of a 2-dimensional regression function in order to determine the displacement as for the intensity tracking technique (Strozzi et al., 2002). Small tracking window sizes are recommended like for the tracking of the speckle pattern. The quality of the results depends on the level of coherence and can be very high (Gray et al., 1998). 4.2 Altimetry The concept of a laser altimeter is quite simple. It sends out a short laser pulse (usually a few nanoseconds long) towards the Earth's surface. Typically, lasers emitting near-infrared or visible electromagnetic radiation are used (see Table 4-1). The laser rays are partly reflected back to the instruments and are detected by a photo-diode. The two-way travel time of the laser pulse is obtained by measuring the time lag between the emission and detection of the laser pulse at the altimeter. The distance (range) to the reflecting surface is calculated, by applying the propagation speed of the laser beam (Pellikka and Rees, 2010). This technique is often called LiDAR (Light Detection And Ranging). However, the propagation speed of the laser rays through the atmosphere is affected by the dry atmosphere and its water vapor content and depends on the wavelength of the laser. Therefore, atmospheric corrections need to be applied. The absolute location of the observed object can be derived if the absolute position of the laser altimeter is known (e.g. by GPS positioning) (Rees, 2006). Airborne altimeters can often measure the range in various directions perpendicular to the flight direction by changing the angle of radiation. This technique is called laser scanning and facilitates the measurement of topographic heights within a corridor along the flight line (Pellikka and Rees, 2010). The spaceborne altimeter GLAS on board the ICESat satellite performed only nadir ranging, however its successor ICESat-2 will be able to obtain multiple measurements in the across-track direction (Abdalati et al., 2010; Schutz et al., 2005). The principle of a radar altimeter is very similar to that of a laser altimeter. The radar signals penetrate through clouds or fog in contrast to the electromagnetic energy pulses of laser altimeters (Massom and Lubin, 2006). Therefore, radar altimeters can provide data in frequently clouded regions (e.g. Antarctic Peninsula). However, the footprints of spaceborne nadir radar altimeters are very large (e.g. Envisat: 1.2-16 km; Brenner et al., 2007) compared to laser altimeter footprints (Table 4-1) due to different radiation characteristics of the electromagnetic wave sources. Hence, these instruments are suitable for monitoring regions with smooth topography such as ice shelves or the central parts of the large ice sheets. The SIRAL radar altimeter of Cryosat-2 is equipped with two antennas allowing the system to work in synthetic aperture interferometric mode (SARIn) (Wingham et al., 2006). Using Cryosat-2 SARIn data and sophisticated post-processing techniques, high quality topographic measurements with footprints in the range of 200 m (personal communications with Veit Helm) can be obtained (Helm et al., 2014). Hence, Cryosat-2 measurements can be applied in areas with sufficiently gentle topography (Section 13.1). In this work, space-borne altimetry data from ICESat (Zwally et al., 2012b), Cryosat-2 (Helm et al., 2014) and airborne laser altimetry data from the Airborne Topographic Mapper (ATM) instrument, acquired during Operation IceBridge surveys (Krabill, 2010), were used for the vertical referencing of DEMs from various sources. 20 Remote sensing of polar ice sheets 4.3 Multispectral satellite remote In addition to SAR imagery, acquisitions from multispectral sensors were analyzed in this work (Table 4-1). Landsat data was used for manual mapping of the lateral boundaries of valley glaciers (trimline) and surface elevation changes were obtained by means of SPOT 5 and Terra satellite data. DEMs can be derived from stereoscopic image pairs (acquisition of the same region from different viewing angles) by applying photogrammetric techniques (Massom and Lubin, 2006). Both satellites, SPOT 5 and Terra, were equipped with sensor pointing in different directions along the flight line (SPOT 5 HRS: 20° towards the rear and front; Terra ASTER: nadir and 27.6° towards the rear, Chanel 3 near-infrared). Processed DEMs were provided by the SPOT 5 stereoscopic survey of Polar Ice: Reference Images and Topographies (SPIRIT) project during the fourth International Polar Year (2007-2009) (Korona et al., 2009) and for Terra ASTER by NASA LP DAAC (NASA LP DAAC, 2007). 4.4 Ground penetrating radar sensing In order to obtain information on ice thickness, a similar concept to that used for altimetry measurements is applied. Radio waves typically in the VHF band (5-300 MHz) are transmitted and received by the ground penetrating radar (GPR). Fresh water ice is nearly transparent for electromagnetic waves at these frequencies (Rees, 2006). Reflections of the emitted radar pulses on the ice surface and the ice-bedrock or ice-water interfaces are detected. The ice thickness is derived from the temporal delay between both echoes and the propagation speed of the radio waves in the media. The beam width of the transmitted signal is quite wide due to the long wavelengths of the radio signals. Therefore, GPR measurements are done in-situ with the instrument located on the ice surface or with aircrafts flying at low altitudes in order to obtain a certain spatial resolution (Pellikka and Rees, 2010). Ice thickness data from the airborne Multichannel Coherent Radar Depth Sounder (MCoRDS) recorded during Operation IceBridge flights (Leuschen et al., 2010) was integrated in this work. The MCoRDS system is a five-channel monostatic radar instrument, operating at a center frequency of 193.9 MHz and a pulse repetition frequency of 9000 Hz. Chirped pules with signal durations of 1,10 or 30 μs can be selected depending on the flight altitude (Shi et al., 2010). The processing of the MCoRDS data was performed at the Center for Remote Sensing of Ice Sheets (CReSIS). 4.5 Glacier mass balance derived from remote sensing data The difference between gains (accumulation) and losses (ablation) of a glacier, observed over a defined time interval, is defined as its mass balance. Annual mass balance is the variation in glacier mass between the same dates in subsequent years. However, annual measurements are often performed at the time of the annual glacier mass minimum. This time span is called the balance year and is not constant, due to variable climatic conditions each year (Benn and Evans, 1998). Mass balances (mass loss) can be converted into contribution to sea level rise (SLR). A common approach is the mass change divided by the product of the fresh water density (1000 kg/m³) and the area of the ocean (362.5x10¹² m²) (Cogley et al., 2011). However, the mass changes of glaciers with floating termini (the vast majority at the polar ice sheets) do not completely contribute to SLR. Only the grounded section is relevant for SLR computations since the floating regions, including the ice shelves, have got a water displacement equivalent to their 21 Remote sensing of polar ice sheets mass (Figure 4-2). Various mass balance measurement methods exist (e.g. flux divergence method, snow line elevation method, hydrologic method, direct measurements, geodetic method, input/output method, etc.). Direct measurements of accumulation are typically done by digging snow pits or drilling cores and analyzing the snowpack and layer thickness. Field observations of ablation are usually done by drilling a network of stakes into the glacier in order to measure the lowering of the surface. Both techniques provide point information of the mass balance and need to be extrapolated for glacier-wide estimates or can be used as input for models. The quality of the glacier-wide mass balance strongly depends on the number and spatial distribution of the sampling points. Direct mass balance observations are very laborious and restricted to the accessible sections of a glacier. Thus, they cannot be carried out over large areas and difficult terrain. Consequently, remote sensing data and methods are ideal tools to monitor the mass balance of the large polar ice sheets. Remote sensing methods to obtain mass balance information include the gravimetric mass balance method, the geodetic mass balance method and the input/output method. The latter two were applied in this work (Fountain et al., 1997; UNESCO, 1970). 4.5.1 Gravimetric mass balance method In July 2003, the Gravity Recovery And Climate Experiment (GRACE) was launched. It consists of two identical satellites flying with an along-track spacing of about 220 km in near-circular orbits. The changes in speed and distance between both satellites, caused by anomalies of the Earth's gravity field, are detected by a highly precise inter-satellite K-Band microwave ranging system (Tapley et al., 2004). Ice mass losses affect the local gravitational field of the Earth. Changes in gravity can also be caused by crustal uplift and mantle flow, and need to be considered in order to derive accurate mass change estimates of the cryosphere (Massom and Lubin, 2006). GRACE is a good tool to derive mass balance estimates of the polar ice sheets. However, it is not possible to obtain mass change measurements at length scales of tens of kilometers, which is the typical size of glacier catchments in the studied region of this work. Moreover, the ice shelves in the study

Figure 4-2: Mass balance scheme of a tidewater glacier. dM: Mass change; SLR: Fraction of dM upstream of the grounding line which contribute to sea level rise

22 Remote sensing of polar ice sheets region disintegrated in 1995; hence, these events are not covered by the GRACE data time series, which started in 2003. Therefore, the gravimetric mass balance method was not applied. 4.5.2 Geodetic mass balance method The geodetic mass balance of a glacier can be obtained by integrating surface elevation change data multiplied by an assumed ice density over its area (Fountain et al., 1997; Haakensen, 1986; Krimmel, 1989). However, in order to obtain ice mass change information of the floating sections of marine and lake terminating glaciers as well as ice shelves, the elevation changes need to be converted into ice thickness changes by considering the hydrostatic equilibrium (Figure 4-2). Information on the elevation change can be determined from repeat altimeter measurements or by differencing of DEMs with different dates. The spatially distributed altimeter measurements need to be interpolated over the glacier area. Thus, DEMs covering the widest parts of the glacier are more appropriate, especially in regions with spare altimeter coverage. The geodetic mass balance method also reveals the spatial variabilities of mass changes. However, the applied ice density can strongly affect the resulting mass balance values and needs to be carefully selected for the individual glacier settings. Moreover, changes in the atmospheric conditions (e.g. accumulation, surface temperature) can lead to temporal variations in the firn densification, which need to be considered in the elevation change measurements, where appropriate (Herron and Langway, 1980; Huss, 2013; Li and Zwally, 2011). 4.5.3 Input/Output mass balance method If the ice thickness is known, the mass balance of a glacier area can be obtained by differencing the mass input on the glacier surface and the ice mass discharge at a flux gate (Cogley et al., 2011; Fountain et al., 1997), located at the lower end of the observed area. This method is also called the “flux gate” or “mass budget” method. It is typically applied at marine or lake terminating glaciers with the flux gate located close to the grounding line (location where the glacier tongue starts to float) position. In contrast to the geodetic method, more variables need to be known or estimated. The ice mass discharge at the flux gate is usually derived from the surface velocities (e.g. SAR offset tracking) and the ice thickness (GPR, modeling), considering the basal friction and velocity distribution in the vertical ice column (Paterson, 2000). Climatic mass balance is applied as the mass input on the surface. It includes the accumulation (surface plus internal accumulation) and ablation of the glacier surface and can be obtained from in-situ measurements or atmospheric modeling (Cogley et al., 2011).

23 Study site

5 Study site

Glaciological changes in one of the fastest changing regions on Earth (Turner et al., 2005), the Antarctic Peninsula, are analyzed in this work. A focus is put especially on the northern section of the peninsula.

Figure 5-1: Overview map of the study region, the Antarctic Peninsula. Surface elevation information from the Bedmap2 dataset (Fretwell et al., 2013). Coastlines and former ice shelf extents from SCAR Antarctic Digital database, version 6.0.

24 Study site 5.1 Geography and climate Antarctica is the southernmost continent, located around the geographic South Pole. It is the fifth- largest continent and extends over about 14 million km². Nearly all of Antarctica (~98%) is covered by ice with a mean thickness of 1937 m. Its ice sheet is the largest store of fresh water on the Earth, containing about 61 % of all fresh water (27 million km³ of ice) (Fretwell et al., 2013; Gleick et al., 1993). Based on climatological and geographical settings, the continent can be separated into three sections: East and West Antarctica divided by the Transantarctic Mountains and the Antarctic Peninsula (Figure 5-1). The Antarctic Peninsula is a heavily glaciated mountain chain, stretching ~1300 km from the Eklund Islands on the west side and Cape Adams on the east side northward to the Antarctic Sound. It covers an area of about 4.2*105 km² (basins 24-27 from Zwally et al., 2012a) and is the northernmost section of Antarctica's mainland, pointing towards South America (Stewart, 2011). The highest peaks of the Antarctic Peninsula reach up to ~2800 m a.s.l. and numerous islands surround the Antarctic Peninsula. The sea around the Antarctic Peninsula is covered by sea ice in winters. The sea ice in the Weddell Sea, east of the Antarctic Peninsula, remains most of the year and only along the north-eastern coast widespread sea ice free areas are frequent in summers. The Bellingshausen Sea on the west coast of the Antarctic Peninsula is widely ice free in summers, with minimum sea ice extents in February/March (Jacobs and Comiso, 1997; Riffenburgh, 2007). The ice sheet of the Antarctic Peninsula is a complex mountainous glacier system and differs strongly from the ice sheets of East and West Antarctica. From the high elevation plateau areas, valley glaciers flow down towards the sea. The outlet glaciers terminate as floating or grounded tidewater glaciers or nourish the numerous ice shelves along the peninsula (Cook et al., 2012). The Larsen Ice Shelves along the east coast, especially the Larsen-C Ice Shelf, are the most prominent. On the west side the Wilkins and George VI ice shelves are the largest ice shelves (Cook and Vaughan, 2010; Rignot et al., 2013). The Antarctic Peninsula has the mildest climate in Antarctica. The warm, maritime climate on the west coast is separated by the Antarctic Peninsula's mountain range from the continental climate on the east side. Therefore, a temperature gradient of 7°C between both sides is observed at similar latitudes (Martin and Peel, 1978). Along the eastern coast, the air flows frequently from the south, due to the low-pressure center in the Weddell Sea and barrier wind effects caused by the topography of the peninsula, deflecting easterly air flow along the southern Weddell See towards the north (Schwerdtfeger, 1984; Turner, 2002). The climate along the west coast is strongly influenced by the circumpolar westerlies flow. The Antarctic Peninsula acts as an orographic barrier for the depressions coming in from the west, which bring relatively mild moist air masses from the South Pacific and the Bellingshausen Sea (Fernandoy et al., 2012). Hence, along the west side and on the plateau region the accumulation rates are highest, with a strong decrease towards the east coast. The climatic mass balance is in general similar to the accumulation and shows a large spatial variability controlled by the topography (Turner, 2002; van Wessem et al., 2016). On the Antarctic Peninsula the mass input rate is continent-wide the highest. The area of the Antarctic Peninsula is only about 3% of the Antarctic Ice Sheet, but its total climatic mass balance amounts to 351 Gt/a, corresponding to ~20% of the total climatic mass balance of Antarctica (van Wessem et al., 2014, 2016). During summers the temperatures on the Antarctic Peninsula are frequently

25 Study site above 0°C and lead to widespread surface melt (Barrand et al., 2013b; Rau and Braun, 2002). About two thirds of Antarctica's meltwater (59 Gt/a) is produced at the Antarctic Peninsula and its surrounding ice shelves (Kuipers Munneke et al., 2012). The westerlies air flow across the Antarctic Peninsula mountain range leads to frequent foehn wind events on the east coast. These events cause strong and sometimes long lasting increases in surface air temperatures by up to 20°C on the lee side of the mountain chain combined with low humidity. Widespread surface melt on the is associated with these foehn wind situations (Cape et al., 2015; Luckman et al., 2014). The most important climatological parameter for glacier mass budget studies is the climatic mass balance. Field observations are sparse at the Antarctic Peninsula and most measurements are performed in the good accessible dry areas and/or over the flat ice shelves (Favier et al., 2013). An updated database of quality controlled climatic mass balance measurements (e.g. ice cores, stake measurements) in Antarctica is available via: http://www-lgge.ujf-grenoble.fr/ServiceObs/SiteWebAntarc/database.php. However, the complex terrain and high spatial variability of climatic parameters make the comparison of climatic mass balance values from different sites or the derivation of climatic mass balance data in unobserved areas difficult. Therefore, computer models are applied in order to obtain region-wide information. Data from the polar version of the Regional Atmospheric Climate MOdel (RACMO) is mostly used by the scientific community to analyze ice sheet mass balances in Antarctica. It comprises the dynamics packages of the High Resolution Limited Area Model (HIRLAM) and the physics package of the European Center for Medium-range Weather Forecasts (ECMWF) Integrated Forecast System (IFS). A multi-layer snow model, a prognostic scheme to compute surface albedo and a snow drift routine are incorporated in the model. It is a hydrostatic model with 40 vertical levels and is forced at the lateral boundaries with ERA-Interim re-analysis data at 6 h temporal resolution (Dee et al., 2011). Two versions of the RACMO2.3 model are available for the Antarctic Peninsula, covering the period 1979-2014. The high-resolution model version (5.5 km grid size, van Wessem et al., 2016) only covers the Antarctic Peninsula, whereas the coarser version (27 km grid size, van Wessem et al., 2014) extents over the whole continent. 5.2 Climatic and glaciological changes The Antarctic Peninsula is a hotspot of global warming. The temperature rise is significantly greater at the Antarctic Peninsula than at the rest of the continent and the global mean. Temperature records from the stations (Faraday/Vernadsky, Bellingshausen, Esperanza, Orcadas) showed a mean atmospheric warming trend of 3.7 ± 1.6 °C/century in the period 1904-2001, which is several times the rate of the global warming trend of 0.6 ± 0.2 °C in the 20th century (Houghton et al., 2001; IPCC, 2007; Vaughan et al., 2003). However, since the late 1990s an absence of the warming trend has been observed at the Antarctic Peninsula, attributed to a gain of cold, east-to- southeasterly wind situations (Turner et al., 2016). Firn and ice core observations revealed increasing climatic mass balance during the last 50 years at the Dolleman Island, and Gomez drill sites (Peel, 1992) and a doubling of accumulation on the western Antarctic Peninsula at the Gomez drill site since 1850 (Thomas et al., 2008). Coincident changes in the adjacent ocean were also reported. The sea ice concentration and cover duration in the Bellingshausen Sea decreased (Jacobs and Comiso, 1997; Smith and 26 Study site

Stammerjohn, 2001; Stammerjohn et al., 2008). A rise in surface summer temperatures of more than 1°C and a strong upper-layer salinification of the sea was observed by Meredith and King (2005), attributed to atmospheric warming and reduced sea ice production. Warming of the sea leads to increased oceanic melting of tidewater glaciers and ice shelves (Cook et al., 2016; Dutrieux et al., 2014; Holland et al., 2010). The ice shelves along the Antarctic Peninsula are strongly influenced by climatic changes. The determined thermal limit of ice shelf viability, the -9°C annual isotherm, propagated south due to the atmospheric warming (Morris and Vaughan, 2003; Vaughan and Doake, 1996), resulting in a significant retreat of the ice shelves around the Antarctic Peninsula within the last century. Since the 1960s, the ice shelves have shrunk by about 28000 km², which corresponds to 18% of the ice shelf area (Cook and Vaughan, 2010). All ice shelves, except the Larsen-D Ice Shelf, showed a recession. Some experienced major retreat events, e.g., the Wilkins Ice shelf in 2008 (Braun et al., 2009; Braun and Humbert, 2009), or disintegrated, e.g., the Larsen-A and Larsen-B ice shelves (e.g. Glasser and Scambos, 2008; Rack and Rott, 2003; Rott et al., 1996; Skvarca et al., 1999). The ice shelves around Antarctica are a safety band (Fürst et al., 2016). Drag forces, caused by the lateral and basal stress of the ice shelf at sidewalls, ice rises and ice rumbles, buttress the ice flow of the ice shelf tributaries (Goldberg et al., 2009; Schoof, 2007). Consequently, former tributary glaciers reacted with accelerated ice flow and further frontal recession to the retreat or disintegration of the ice shelves. At the Larsen-A and Prince-Gustav Ice Shelf tributaries sped-up and a retreat of the glaciers as a consequence of ice shelf disintegration in 1995 were reported by various studies (e.g. De Angelis and Skvarca, 2003; Glasser et al., 2011; Rack and Rott, 2003; Rott et al., 2014, 2007, 2002). The Larsen-B tributary glaciers lost about 270 km² of grounded ice after the ice shelf collapse in 2002 and showed an acceleration of ice flow by up to 4-5 times, followed by a subsequent deceleration (Rott et al., 2011; Wuite et al., 2015). Rignot et al. (2004) reported even higher accelerations of up to 8.7 times of the pre-collapse values. Moreover, the authors measured high glacier thinning rates in the range of 10 and 52 m/a in 2003 as a consequence of the increased ice discharge. Comparable values of surface lowering were also observed by Berthier et al. (2012) and Shuman et al. (2011). Various groups of authors have calculated the mass budget of the former ice shelf tributaries. At the Larsen-B embayment, geodetic mass balances of -9.0±2.1 G/a and -8.8±1.6 Gt/a were observed based on differencing of DEMs and altimetry data in the periods 2006-2010/11 and 2001/02-2006, respectively (Berthier et al., 2012; Shuman et al., 2011). Mass losses, estimated by means of the changes in ice discharge of the glaciers (Input/Output method), of -21.9±6.6 Gt/a and -4.3±1.6 Gt/a were reported by Rignot et al. (2004) and Rott et al. (2011), respectively. Scambos et al. (2014) measured geodetic mass balances of -4.5 Gt/a and -8.0 Gt/a in the period 2003-2008 at the Larsen-A and Larsen-B tributaries, respectively. At the Prince Gustav tributaries, they found an imbalance of -2.7 Gt/a. Along the Nordenskjöld coast (Prince Gustav and Larsen-A tributaries), Rott et al. (2014) calculated a mass change rate of -4.2±0.4 Gt/a based on TanDEM-X satellite measurements in the period 2011-2013. The results of the various studies vary significantly and indicate the large spatial and temporal variability of the mass loss. Marine terminating glaciers along the Antarctic Peninsula also showed considerable variations. About 90% of the glaciers north of 70°S have retreated since the 1940s (Cook et al., 2005, 2014). 27 Study site

The highest retreat rates were found on the southwestern coast and were attributed to rising mid- depth ocean temperatures in this region (Cook et al., 2016). Furthermore, the ice flow speeds of glaciers along the Antarctic Peninsula's west coast accelerated by ~12% between 1992 and 2005 (Pritchard and Vaughan, 2007), which is attributed to frontal thinning. This assumption is supported by an observed mean surface lowering rate of 0.28±0.03 m/a close to the terminus of 12 glaciers on the western Antarctic Peninsula between the 1960s and 2000s (Kunz et al., 2012). Moreover, Helm et al. (2014) and Wouters et al. (2015) obtained significant surface lowering of glaciers on the southern Antarctic Peninsula, which Wouters et al. (2015) associated with the intrusion of warm Circumpolar Deep Water onto the continental shelf. The observed changes in ice dynamics suggest that the Antarctic Peninsula Ice Sheet has negative mass balances and contributes to the global sea level rise. Sasgen et al. (2013) presented a mass balance of -26±3 Gt/a for the northern Antarctic Peninsula (area north of Larsen- D and George VI ice shelves) in the time interval 2003-2012, obtained from GRACE data. A similar trend was derived from GRACE measurements by Groh and Horwath (2016) of -20.6±5.9 Gt/a in the period 2002-2016 (data available via https://data1.geo.tu-dresden.de/ais_gmb/). The islands around the Antarctic Peninsula were included in Sasgen et al. (2013). Gardner et al. (2013) obtained a mass change rate of -7.3±4.0 Gt/a (2003-2009) from ICESat data for the Antarctic Peninsula's peripheral islands, explaining the differences between both GRACE studies. Moreover, both studies show nearly balanced mass budgets for the southern Antarctic Peninsula. Shepherd et al. (2012) performed a reconciled estimate of the polar ice sheet mass balance. The authors reported a mass change of -380±266 Gt on the Antarctic Peninsula in the period 1992-2011(- 20±14 Gt/a). This corresponds to nearly one third of the continent-wide mass loss. The imbalance trend increased within the study period (-29±12 Gt/a in 2000-2011). Scambos et al. (2014) computed a mass change rate of -24.9±7.8 Gt/a at regions north of 66° S between 2003 and 2008, with major mass losses on the east coast. Both studies reflect the dramatic mass loss rates of tributary glaciers after ice shelf collapse along the north-eastern Antarctic Peninsula. However, reduced mass loss rates at the northern Antarctic Peninsula of -13±18 Gt/a were derived from CryoSat-2 between 2010 and 2013 (McMillan et al., 2014). Barrand et al. (2013a) modeled the mass budget of Antarctic Peninsula Ice Sheet until 2200. A net sea level contribution of 7-11 mm sea level equivalent (~2540-3990 Gt) by 2100 and 10-25 mm (~3600-9060 Gt) by 2200 was estimated. Region-wide mass budget estimates are strongly influenced by the huge temporal and spatial variability of the mass balance. Moreover, each method is based on specific assumptions and has its limitations. Consequently, a comparison of different mass balance estimates needs to be carefully discussed.

28 Motivations and Objectives

6 Motivations and Objectives

The Antarctic Ice Sheet has clearly lost ice within the past decades. Moreover, the imbalance has even increased (IPCC, 2014). The reported regional and local mass balance estimates on the Antarctic Peninsula (Section 5.2) indicate that this region is significantly contributing to global SLR. However, the results vary strongly, indicating the large spatial and temporal variability of mass balance along the Antarctic Peninsula. Region-wide mass budget estimates are considerably influenced by this variability. Therefore, glaciological changes on small spatial and temporal scales need to be quantified. At the northern Antarctic Peninsula, the disintegration of the ice shelves (Prince-Gustav, Larsen-A/B) triggered major mass loss of the tributary glaciers (e.g. Berthier et al., 2012; Rack and Rott, 2004). Similar processes also occurred at the Wordie Ice Shelf located on the western Antarctic Peninsula (e.g. Doake and Vaughan, 1991; Wendt et al., 2010). The thinning of ice-shelves, especially along the Antarctic Peninsula and West Antarctica, was reported by Paolo et al. (2015). A large rift has grown at the Larsen-C Ice Shelf and numerous rifts were observed at the SCAR Inlet Ice Shelf. These findings are warning signs of potential ice shelf destabilization and break up (Jansen et al., 2015; Khazendar et al., 2015; Wuite et al., 2015). The observed acceleration and surface lowering of (e.g. Flament and Rémy, 2012; Rignot, 2008) is assumed to be forced by a reduced buttressing by its ice shelf (Dutrieux et al., 2014). The Pine Island Ice Shelf has significantly retreated and thinned. These changes were attributed to a change in oceanic conditions (e.g. Corr et al., 2001; Jacobs et al., 2011). A collapse of Pine Island Glacier could lead to a destabilization of the entire West Antarctic Ice Sheet, resulting in a dramatic contribution to SLR (Hughes, 1981; Pearce, 2007). Moreover, further changes in climatic and oceanic conditions can lead to further thinning and mass loss of ice shelves. Hellmer et al. (2012) have projected a change of ocean currents in the 21st century that could result in dramatic mass losses of the larger Filchner-Ronne Ice Shelf (southern Weddell Sea). The Filchner-Ronne Ice Shelf buttresses some of the largest ice streams in Antarctica, draining major parts of its ice sheet. In order to better estimate the potential consequences and risks of ice shelf collapses, profound knowledge of the processes is necessary. Detailed analysis of previous events will help to improve the information and understanding of the ongoing processes. Therefore, spatially and temporally detailed time series studies on the tributary glaciers' response to the disintegration of the Larsen-A and Prince-Gustav ice shelves were carried out. Some information on glacier mass balances or dynamics of these glaciers do exist (Section 5.2), but they just cover specific periods or focus on certain aspects of glacier change. Moreover, they were derived with varying methods, making the comparison and analysis of temporal changes difficult. Thus, the aim of this work is to observe the changes in ice dynamics and mass balances with consistent methods by considering all available datasets. Moreover, the integration of newly available datasets in this region (e.g. ice thickness information by Huss and Farinotti (2014) and climatic modeling data by van Wessem et al. (2016)) facilitates an upgraded analysis. Along the west coast of the Antarctic Peninsula, glaciological changes are ongoing as well (Section 5.2). Published studies often just cover one aspect or have partial spatial coverage (e.g. Kunz et al., 2012; Pritchard and Vaughan, 2007). Moreover, the high climatic mass balance values on the 29 Motivations and Objectives west side (Turner, 2002; van Wessem et al., 2016) and the increase in accumulation on the Antarctic Peninsula (Peel, 1992; Thomas et al., 2008) might balance the mass loss of the glaciers. To tackle these ambiguities, comprehensive observations are needed (Kunz et al., 2012). Hence, a comprehensive time series analysis of the glacier changes along the northern Antarctic Peninsula, covering the east and west coast, was conducted. The aim was to obtain information on the different glaciological variations in this region and their spatial distribution. Moreover, correlations between glacier changes and atmospheric, oceanic and glaciological forcing should be analyzed to facilitate a better understanding of their relationship. Additionally, an updated mass balance estimation and its temporal change of the study region is desirable in order to provide information for updated continent-wide mass balance computations, such as the Ice sheet Mass Balance Inter- comparison Exercise (IMBIE). The TanDEM-X Mission (launched in 2010) provides a unique instrument to observe glaciological changes (Section 4.1.1.1). On the northern Antarctic Peninsula, numerous “Super Test Sites” are defined in the mission plan. Bistatic InSAR data is acquired regularly at these sites. Dense TanDEM-X time series data was integrated in all studies in order to assess its capabilities to monitor glaciological changes and to infer short term fluctuations in combination with other sensors. The intended major objectives of this work are: 1. Time series analyses of surface velocity and glacier area changes 2. Temporal trends of surface elevation changes 3. Variabilities of grounding line position 4. Glacier mass balance estimations for different time periods 5. Glacier mass balance derived with different methods 6. Spatial distribution of changes in ice dynamics 7. Observation of short term or seasonal changes of ice dynamics 8. Testing the potential of TanDEM-X data to monitor the polar ice sheets

30 Motivations and Objectives

6.1 Contributions to scientific papers Three specific studies were carried out to investigate the target region to accomplish the proposed objectives. The outcomes from the first and second study are published in peer-reviewed journals. A paper about the third study is ready to be submitted. The author's contribution and a short summary of each study are listed below: Study 1: Title: Changes in ice dynamics, elevation and mass discharge of Dinsmoor–Bombardier– Edgeworth glacier system, Antarctic Peninsula Authors: Seehaus, T., Marinsek, S., Helm, V., Skvarca, P., Braun, M. Journal: Earth and Planetary Science Letters (2015) Short summary: A detailed study on the reaction of the Dinsmoor-Bombardier-Edgeworth glacier system to the disintegration of the Larsen-A Ice Shelf in 1995 was performed. Dense time series of glacier velocities, area change, ice mass flux and surface elevation change were derived from remote sensing data in the period 1992-2014. Mass balances and the contribution to SLR of the glacier system are calculated for different periods, using the Input/Output and geodetic mass balance method and considering different ice discharge and fjord depth scenarios. A rapid and dramatic increase in ice flow after the ice shelf break up was observed with a subsequent deceleration. The surface lowering trend showed an exponential decay and the mass loss was significantly higher in the first decade after the disintegration of Larsen-A compared to the most recent estimates. The glacier system was still contributing to SLR in 2014, however, it slowly adapts to the new boundary conditions. Objectives No.: 1-5, 7, 8 Own contribution: design of the study, processing of remote sensing data, analysis, creation of figures and tables, writing of the manuscript Study 2: Title: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series Authors: Seehaus, T., Marinsek, S., Skvarca, P., van Wessem, J.M., Reijmer, C.H., Seco, J.L., Braun, M. Journal: Frontiers in Earth Science (2016) Short summary: The response of the glaciers at the Sjögren Inlet to the retreat and breakup of the Prince-Gustav Ice Shelf in the early 1990's was analyzed in the period 1993- 2014. Changes in ice surface velocities, mass flux, glacier front position and ice surface elevation were investigated. Imbalances of the glaciers were estimated for different time periods. Acceleration in glacier flow followed by a slow-down, as well as significant frontal retreat and surface lowering were found as a consequence of the ice shelf break up. High mass loss rates were observed in the 1990's, which clearly decreased afterwards. Differences in the dynamic response of Sjögren Inlet glaciers and the Dinsmoor- Bombardier-Edgeworth glacier system to the same forcing were attributed to the individual glacier settings and fjord geometries.

31 Motivations and Objectives

Objectives No.: 1-4, 7, 8 Own contribution: design of the study, processing of remote sensing data, analysis, creation of figures and tables, writing of the manuscript Study 3: Title: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 Authors: Seehaus, T., Cook, A., Barbosa, A., van Wessem, J.M., Reijmer, C.H., Braun M. Journal: to be submitted to The Cryosphere Short summary: The changes in glacier flow and area of 74 major glaciers along the northern Antarctic Peninsula were analyzed in the period 1992-2014. The study region was divided into three sections, considering the different climatic, oceanic and glaciological conditions. Ice discharge was obtained for the periods 2010-2014 and 1992-1996. Mass budgets of the sections were calculated by means of climatic mass balance modeling results. Glaciers affected by ice shelf disintegration showed similar behavior. The glaciers in the north-eastern section retreated and their ice flow decreased in general. Along the western side, the glaciers showed a heterogeneous pattern of area and velocity changes. By applying a cluster analysis, changes in ice dynamics could be attributed to their geometries. Glaciers on the east coast showed negative mass balances in the 1990's and nearly balanced mass budgets in recent years. Very high accumulation rates were found along the west coast, leading to positive mass balances of the whole study region. Objectives No.: 1, 4, 6-8 Own contribution: design of the study, processing of surface velocity fields, update of glacier area change inventory, analysis, creation of figures and tables, writing of the manuscript

32 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

7 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

Seehaus, Thorsten1; Marinsek, Sebastián2,3; Helm, Veit4, Skvarca, Pedro5; Braun, Matthias1

1 Institut für Geographie, Universität Erlangen-Nürnberg, Wetterkreuz 15, D-91058 Erlangen, Germany 2 Instituto Antártico Argentino, Balcarce 290, C1064AAF, Buenos Aires, Argentina 3 Universidad Tecnológica Nacional, Facultad Regional Buenos Aires, Medrano 951, C1179AAQ, Buenos Aires, Argentina 4 Alfred-Wegener-Institut für Polar- und Meeresforschung, Am Alten Hafen 26, D-27568 Bremerhaven, Germany 5 Glaciarium, Museo del Hielo Patagónico, El Calafate 9405, Prov. Santa Cruz, Argentina Published in Earth and Planetary Science Letters 427 (2015)

Abstract

The northern Antarctic Peninsula is one of the fastest changing regions on Earth. The disintegration of the Larsen-A Ice Shelf in 1995 caused tributary glaciers to adjust by speeding up, surface lowering, and overall increased ice-mass discharge. In this study, we investigate the temporal variation of these changes at the Dinsmoor-Bombardier-Edgeworth glacier system by analyzing dense time series from various spaceborne and airborne earth observation missions. Precollapse ice shelf conditions and subsequent adjustments through 2014 were covered. Our results show a response of the glacier system some months after the breakup, reaching maximum surface velocities at the glacier front of up to 8.8 m/d in 1999 and a subsequent decrease to ~1.5 m/d in 2014. Using a dense time series of interferometrically derived TanDEM-X digital elevation models and photogrammetric data, an exponential function was fitted for the decrease in surface elevation. Elevation changes in areas below 1000 m a.s.l. amounted to at least 130±15 m between 1995 and 2014, with change rates of ~3.15 m/a between 2003 and 2008. Current change rates (2010–2014) are in the range of 1.7 m/a. Mass imbalances were computed with different scenarios of boundary conditions. The most plausible results amount to −40.7±3.9 Gt. The contribution to sea level rise was estimated to be 18.8±1.8 Gt, corresponding to a 0.052±0.005 mm sea level equivalent, for the period 1995–2014. Our analysis and scenario considerations revealed that major uncertainties still exist due to insufficiently accurate ice-thickness information. The second largest uncertainty in the computations was the glacier surface mass balance, which is still poorly known. Our time series analysis facilitates an improved comparison with GRACE data and as input to modeling of glacio-isostatic uplift in this region. The study contributed to a better understanding of how glacier systems adjust to ice shelf disintegration. Keywords: Larsen-A Ice Shelf; remote sensing; glacier change; contribution to sea level rise; glacier mass balance; sensitivity analysis

33 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula 7.1 Introduction Climate conditions along the northern Antarctic Peninsula (AP) have undergone remarkable changes during the last decades. Together with West Antarctica, the AP contributes to the current mass loss of the Antarctic Ice Sheet (Pritchard et al., 2012; Shepherd et al., 2012). A significant increase in surface air temperatures was reported for the AP (Skvarca et al., 1998) and changes in the precipitation pattern are also known (Turner, 2002). Widespread glacier retreat has been observed by Cook et al. (2005) along the AP over the past decades. Pritchard and Vaughan (2007) reported accelerated glacier flow along the west coast of the AP, caused most likely by dynamic thinning as a consequence of increased surface and basal melt. Kunz et al. (2012) found a surface lowering at low altitudes on several western AP glaciers. The most dramatic changes were clearly the disintegrations and breakups of several ice shelves (Prince Gustav Channel, Larsen Inlet, Larsen-A/B, Wordie, Wilkins, Jones, Muller) on both coasts of the AP since the late 1980s (Cook and Vaughan, 2010). The buttressing effect and subsequent acceleration of tributary glaciers were documented by Rott et al. (2002) as well as quantified by various other groups (e.g. De Angelis and Skvarca, 2003; Rack and Rott, 2004, 2003; Rignot et al., 2004; Rott et al., 2007). Using repeated satellite acquisitions, Berthier et al. (2012) and Scambos et al. (2004) showed significant glacier surface lowering and speedup. Rott et al. (2011) provided a comprehensive imbalance computation of the former Larsen-B tributary glaciers, although the ice-thickness was unknown for many areas. They used high-resolution TerraSAR-X imagery and data from the ERS mission to compute imbalances via the flux gate (input/output method). Scambos et al. (2014) provided an updated mass balance estimate for the northern AP, using repeat photogrammetric elevation measurements. Rott et al. (2014) used bitemporal TanDEM-X data to reveal the mass loss of northeastern AP (glaciers draining into Prince Gustav Channel, Larsen Inlet and Larsen-A embayments) for the years 2011– 2013. An integrated assessment of glacier mass balance for the entire AP of −20±14 Gt/a (1992– 2011) was provided by Shepherd et al. (2012). However, their uncertainties amounted to up to 70% of the signal. Imbalance computations often still lack a clear presentation of assumptions made, existing data voids, and respective sensitivities considering different scenarios. Only a few attempts have been made to exploit full archives of various satellite and airborne missions (SAR, optical, altimetry, and ground penetrating radar) in this region to study the reaction of tributary glaciers to ice shelf breakup and disintegration in high temporal resolution. The response time of glaciers to the new boundary conditions as well as the duration of such a reaction have not yet been well studied. At , increased ice-flow velocity and continued thinning were found more than 30 years after disintegration of Wordie Ice Shelf (Rignot et al., 2005; Wendt et al., 2010). In this study, we analyzed the archives of different Synthetic Aperture Radar (SAR) missions (ERS- 1/2 AMI SAR, ENVISAT ASAR, RADARSAT-1, ALOS PALSAR, TerraSAR-X, TanDEM-X) to monitor surface velocity changes of three major tributaries of the former Larsen-A Ice Shelf. The data set covers acquisitions before the ice shelf breakup in 1995 and reaches until 2014. The velocity data was analyzed in conjunction with changes in ice-front position. Surface elevation changes were derived from digital elevation models (DEM), using Terra ASTER, SPOT SPIRIT, and the bistatic

34 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula TanDEM-X mission combined with altimetry data (ICESat GLAS, CryoSat-2 SIRAL-2, NASA Operation IceBridge). Ice-thickness measurements from NASA Operation IceBridge and bedrock elevation models from Huss and Farinotti (2014) and SCAR Bedmap2 (Fretwell et al., 2013) were integrated to estimate variations of ice mass flux and frontal retreat at the Dinsmoor-Bombardier- Edgeworth glacier system (DBE) over the time period 1995–2014. Finally, we provide updated imbalance computations for the glacier system, including an error estimation and sensitivity analysis, resulting from the various data sets used, methods applied, and assumptions made. In the terminology used, we refer to the UNESCO glossary of glacier surface mass balance (Cogley et al., 2011). 7.2 Study site

Figure 7-1: Panels a and b: Location of Dinsmoor–Bombardier–Edgeworth Glacier at the Antarctic Peninsula and Antarctica. Map base: Landsat LIMA Mosaic © USGS and SCAR Antarctic Digital Database, version 6.0. Panel c: Surface velocity field and glacier front variation of DBE. Surface velocities were derived from TerraSAR-X acquisitions (August 4 and 15, 2012). Colored lines: Changes of glacier front position picked from SAR intensity images; orange line: Operation IceBridge (OIB) Airborne Topographic Mapper (ATM) profile (Track ID = 0, Nadir) from November 14, 2011; bright green line: glacier catchment from SCAR Antarctic Digital Database, version 6.0; background: Landsat LIMA Mosaic © USGS (for a hypsometric curve of glacier catchment see supplemental material Fig. 13-7).

The DBE tidewater glacier system comprises three large glaciers, jointly forming the northernmost former tributary of the Larsen-A Ice Shelf. It is located between Sobral Peninsula and Fothergill Point. The location and retreat states of DBE since 1995 are shown in Fig. 7-1. The catchment area of DBE is about 610 km² (2014) and its elevation ranges from sea level to about 2200 m.

35 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula Gentle surface slopes at the glacier front turn to a steep elevation step towards the AP plateau. The region is influenced by the Weddell Gyre as well as Westerlies ranging over the AP plateau. The latter cause a decrease of precipitation from the plateau to the east coast. Frequent foehn- type wind events occur (Marshall et al., 2006; Orr et al., 2004), leading to clear sky conditions at the east coast with considerably higher surface air temperatures than on the west coast of the AP. They are associated with high wind speeds (northwesterly to westerly), surface melt, and melt ponding at low elevations. The embayment of the former ice shelf is frequently ice-covered, and consistent summer sea ice coverage in the bay of DBE has been observed since 2009 (by survey flights from Marambio base and satellite imagery). 7.3 Data

Table 7-1: Overview of sensors and their main characteristics used in this study. SAR sensors: Intensity tracking parameters are given in pixels (p) in slant range geometry. The resolution of SAR intensity images used for glacier front mapping is provided in ground range geometry.

A Synthetic Aperture Radar (SAR) image time series was analyzed to study the changes in surface velocities and glacier extent of DBE between 1992 and 2014. Tab. 7-1 shows the specifications and epochs of the SAR sensors used. The period covers pre-ice-shelf-collapse to current

36 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula conditions. The refined ASTER Global DEM of the Antarctic Peninsula (AP-DEM) from Cook et al. (2012) was taken as elevation reference for geocoding and orthorectification of all SAR data. Bistatic interferometric SAR data has been regularly acquired along the AP since 2011, because the Larsen-A and B region had been identified as “super test sites” for the TanDEM-X (TDX) satellite mission. Therefore, DBE was covered with a TDX repeat coverage once or twice a month. The bistatic SAR acquisitions are not subject to temporal decorrelation by surface changes and hence enable regular interferometric derivations of DEMs. For earlier time steps, stereoscopic DEMs from the level-3 standard NASA LP DAAC processing of Terra ASTER (AST14DMO) and the SPOT SPIRIT project (Korona et al., 2009) were integrated (Tab. 7-1). Five DEMs from these multispectral sensors with no or very low cloud cover and 27 TDX DEMs were selected. More TanDEM-X DEMs were processed, but DEMs with phase unwrapping errors were discarded. Altimeter data from the NASA ICESat Geoscience Laser Altimeter System (GLAS) instrument (GLA06 L1B Global Elevation Data Version 33, Zwally et al., 2012), CryoSat-2 SIRAL-2 (Helm et al., 2014) and NASA Operation IceBridge (OIB) Airborne Topographic Mapper (ATM) (Krabill, 2010) were used for referencing and accuracy assessments of the DEMs. Coverage by all DEMs existed only on the mid to lower parts of Bombardier and Edgeworth Glacier. Dinsmoor Glacier and the upper part of Bombardier Glacier were not covered by the ASTER DEM in 2001. The 2006 SPOT SPIRIT DEM did not cover the upper parts of Dinsmoor and Bombardier Glacier, and clouds in the upper part of Edgeworth Glacier affected the 2006 and 2008 ASTER DEMs. The TDX DEMs have only partial or low quality elevation data close to the steep valley side walls. This is caused by layover, shadowing, and foreshortening effects. Ice-thickness information was available from three different sources: (t1) Bedmap2, a gridded Circum-Antarctic data set (Fretwell et al., 2013), (t2) one flight line with the OIB Multichannel Coherent Radar Depth Sounder (MCoRDS) (Leuschen et al., 2010) on November 14, 2011, and (t3) a modeled bedrock elevation grid (Huss and Farinotti, 2014). The OIB MCoRDS data was included in the generation of both gridded bedrock data sets. At our study region, the thin ice model was applied for the Bedmap2 processing. The bedrock grid from Huss and Farinotti (2014) used various ice-thickness data sets for parameter adjustment and validation as well as surface velocities and surface mass balance (1979–2010) as input. The latter most likely represents the best ice-thickness data set currently available for the region. The deficiencies of the Bedmap2 data set are expressed by the fact that it shows bedrock elevations that are higher than the ATM surface elevations in wide parts at Dinsmoor and Bombardier Glacier (Supplemental material, Figure 13-6).

For the surface mass balance (bsfc) of the catchment, an upper and lower bound approximation of bsfc(min) = 1080 kg/(m²a) and bsfc(max) = 1720 kg/(m²a) were defined based on literature data (Supplemental material, Section Error: Reference source not found-S1). The mean annual surface air temperature records (1993–2014) from the two closest Argentine stations, Marambio and Matienzo, were used for the interpretation of the climate conditions.

37 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula 7.4 Methods 7.4.1 Glacier extent Changes in the glacier front positions of DBE after ice shelf disintegration were mapped by using SAR intensity imagery. Single-look complex SAR data was multilooked, geocoded, and orthorectified to generate intensity images (see Tab. 7-1 for details on grid resolution). The glacier front was manually picked, and a quality factor ranging from 1 to 6 was assigned. Image contrast, spatial resolution, and surface structure on the glacier and on the ice mélange in the glacier fjord influenced the precision of the front detection. The quality factors and their related accuracy values perpendicular to the picked front position are shown in Tab. 7-2. Data with a quality factor of 6 were discarded from any further analysis. Temporal variations in the glacier area (Sr) were calculated from the shift of the glacier front position relative to the precollapse grounding line (GL) position in

1995 (Rack and Rott, 2004). The error for the area computations (σSr) resulted from the accuracy of the respective front multiplied by its length. Table 7-2: Quality factors of manual glacier front detection and the assigned horizontal accuracies σf (perpendicular to estimated glacier front position).

7.4.2 Surface velocities We derived ice surface velocity fields to investigate variations in the ice flow. An intensity offset tracking algorithm was applied on time series of coregistered single look complex SAR images (Strozzi et al., 2002). This technique uses a moving window algorithm to calculate the azimuth and slant range displacements of surface structures. It requires the definition of a correlation patch size and a step size (the distance between the centers of the patches). Estimates of low confidence were removed by applying a threshold on the signal-to-noise ratio of the normalized cross- correlation function, which was used as a quality indicator. The tracking processing was done within the GAMMA Remote Sensing software. Tracking parameters were adjusted to the sensor specification and acquisition intervals (max. 92 days) as well as to the expected displacements (Tab. 7-1). Errors were computed for each scene, considering overall scene-to-scene coregistration performance, time interval, and image resolution in ground range. Details of the postprocessing and error estimation of the obtained tracking results are given in the supplemental material in Section Error: Reference source not found-S2. 7.4.3 Surface elevations An average elevation change (dh) for the lower area of the glacier system was calculated from all DEMs. The approximate recent position of the grounding line was determined according to a prominent elevation change pattern between the grounded and floating ice by differencing the DEMs from 2003 and 2014. DEMs were interferometrically derived from the regular bistatic TanDEM-X CoSSC data (DLR-IMF et al., 2012) available since 2011. The resulting TDX DEMs were vertically referenced by means of OIB ATM, and CryoSat-2 data. The stereoscopic DEMs from ASTER and SPOT were vertically

38 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula referenced to ICESat GLAS altimeter profiles. The processing of the TDX DEMs and the quality assessment for all DEMs is described in the supplemental material (Sections Error: Reference source not found-S3 and S4). The glacier surface elevation prior the Larsen-A Ice Shelf collapse in 1995 was assessed by mapping the lateral glacier boundary (trimline) on a Landsat 5 image acquired March 1, 1986. Its elevation was extracted from the SPOT SPIRIT DEM in 2006. This DEM is assumed to have the best horizontal and vertical accuracy on the steep valley sides among the available DEMs (Berthier et al., 2012). The elevation information along the former trimline was manually interpolated over the glacier area by separating the glacier area into several flat subsets. The average elevation values of the related trimline sections were assigned to each subset. However, a reasonable ice elevation estimate was only retrieved on the lower flat parts of DBE by applying this method. We refer to this data set as DEM95. 7.4.4 Ice mass flux To calculated the ice discharge of DBE, the flux gate (FG) was defined along the OIB profile (orange line, Figure 7-1), which is close to our estimated grounding line position (2003–2014, Section 7.5.3). The ice flux (F) through the FG was calculated with the Bedmap2 and Huss and Farinotti bedrock elevation data sets, since the MCoRDS sensor only revealed data at Edgeworth Glacier. Details of the ice flux computation can be found in the supplemental material in Section Error: Reference source not found-S7. 7.4.5 Mass balance To estimate the total mass loss (ΔM) of DBE, we considered both the geodetic (e.g., Scambos et al. 2014) and the input/output approach (e.g., Rott et al. 2011), using different scenarios (surface mass balance bsfc, ice-thickness ti, and ice flux F). For a better understanding of the complex computations, the two approaches and individual components are illustrated in the schematic diagram of Fig. 7-2. We calculated ΔM for three different time intervals:  1995–2003 the period directly after the breakup of Larsen-A (Fig. 7-2a, b, c)  2003–2014 the period with DEM coverage and temporal detailed ice flux data (Fig. 7-2d, e, f)  1995–2014 the whole period since the disintegration of Larsen-A Ice Shelf (Fig. 7-2g, h, I) For each period, the changes of the glacier extent are show in Fig. 7-2. Only changes upstream of the GL position in 1995 were considered. The first approach used the geodetic mass balance method (Fig. 7-2b, e, h). It was computed by g ΔM g= A sl+ A f (5)

In the second approach, we applied the input/output method (Fig. 7-2c, f, i) but in a slightly adapted IO way. Mass loss downstream of the FG (Asl + Af) was estimated using the geodetic mass balance method, IO ΔM IO=MQ+ Asl +A f (6)

39 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula The following mass change components had to be determined:

 Asl by surface lowering

 Af by change in front position

 MQ by flux imbalance (from Section 7.5.4)

Ice mass change derived from surface lowering (Asl) was calculated by DEM differencing on subset area S1 for the DEMs in 2003 and 2014, and subset area S3 (Supplemental material, Figure 13-3) for the DEM95. Scambos et al. (2014) reported an insignificant surface lowering rate of −0.23 m/a (2003–2008) on the plateau regions (>1000 m a.s.l.); consequently, only areas below were included in the estimation of Asl. Zones that were not covered by S1 or S3 were filled by using the elevation change hypsometric curves (Supplemental material, Figure 13-4) and the AP-DEM heights. On floating sections of DBE, the mass loss was derived from the surface lowering and the hydrostatic equilibrium assumption. A density ratio between ice and seawater of 0.87 was applied. Between 1995 and 2003, a linear retreat of the GL and no further retreat of the GL after 2003 were considered in the calculations.

Figure 7-2: Schematic diagram of mass balance calculations for different time intervals (columns) and approaches (rows). Numbers indicate different years/observation periods for glacier surface (S), front position (F), flux gate (FG) and grounding line (GL). No change in grounding line position was assumed for the period 2003–2014. Top row: Glacier extent at the beginning (dashed line) and at the end (solid thick line) of each study period. Central row: Ice mass change derived from the geodetic mass balance method. Lower row: Ice mass change derived by the input–output method (adapted). Mass change variables: Af – by shifting of ice front position; Asl – derived from surface elevation change; MQ – flux imbalance upstream of the flux gate; SLR: Fractions of mass losses which contributed to sea level rise. 40 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

The ice mass loss by retreat of the front position behind the GL in 1995 (Af) was determined from the ice-thickness (ti) and the area of the retreated ice section. For the period 1995–2003, the surface elevation h95 was extracted from DEM95, but no reasonable bedrock elevation data was obtained from the bedrock elevation grids (Bedmap2: minimum = −244 m, mean = −33 m/Huss and Farinotti: minimum = −401 m, mean = −117 m but covers the retreat area only partly).

Therefore, the upper bound of ti was assessed by taking h95 as the freeboard height and calculating ti using the hydrostatic equilibrium assumption. The lower bound of ti was approximated by assuming the same bedrock elevation at the retreat area found for the interpolated MCoRDS profile on Edgeworth Glacier (Supplemental material, Figure 13-6). DBE readvanced between 2003 and 2014. This section was downstream of the GL. Hence, its ice-thickness and mass was derived from the freeboard height. Not all components of ΔM result in a contribution to sea level rise (SLR: striped segments in Fig. 7- 2) and neither could all components be computed with the same precision. The SLR contribution of DBE (ΔMslr) was calculated for both methods and under different constraints. For the geodetic method ΔMslr was calculated by: g slr g ΔM slr=A f + Asl slr (7) and for the flux gate approach by: IO slr IO ΔM slr=A f +M Q+ A sl slr (8)

g slr IO slr where Asl is the fraction of ice mass loss by surface lowering upstream the GL and Asl between the GL and the flux gate, considering the retreat of the GL in the period 1995–2003. slr Taking into account the different densities of seawater and ice, Af is Af (1995–2003) minus its slr volume below sea level times the sea water density. Consequently, Af is zero for the upper bound approximation of Af. 7.5 Results 7.5.1 Glacier extent After the disintegration of the Larsen-A Ice Shelf in 1995, the glacier front started to retreat rapidly in 1996. It retreated behind the former grounding line showing its minimum extent in the image from March 14, 2007, corresponding to a previously grounded ice area loss of 58.6±6.6 km². In Fig.

7-3b, the temporal changes of the glacier area (Sr) are plotted. The clear trend of frontal retreat changed in 2003 and the glacier started to advance again. The readvance was interrupted by at least six significant calving events, which all occurred in austral summers (Jan–Feb. 2005, 2006, 2007, 2010, 2011, 2012; retreat area respectively: 11 km², 14 km², 10 km², 4 km², 11 km², 3 km²).

In March 2014, Sr amounted to 37.5±4.8 km².

41 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

Figure 7-3: Temporal changes of ice dynamics of DBE between 1993 and 2014. Panel a: Surface velocities form SAR intensity tracking; dots denote the median of a profile measured 500 m behind the glacier front position at the respective date; diamonds denote the velocity at the center of flux gate on Edgeworth Glacier (Fig. 7-1). Error bars are the same for both data sets and small error bars are hidden by the plot symbols. *For scaling reasons the maximum measured velocity of 8.8 m/d in 1999 was not plotted. Panel b: Changes in glacier area due to frontal retreat relative to grounding line position in 1995. Panel c: Median surface elevation changes on DBE relative to the Aster DEM in 2003. Blue dashed line: Temporal extrapolation of surface lowering based on subset area 1 data (see Fig. 7-4); pink dashed line: Estimated ice surface elevation prior to ice shelf disintegration in 1995 with an expected accuracy of ± 10 m (pink dotted line). Inset: Zoom on the elevation change derived from TDX DEMs (2011-2014). Panel d: Mean annual surface air temperature at the Argentine stations on the east side of the AP. Approximate position, distance of stations to DBE and data provider: Marambio: 64°14′S 56°37′W, 196 m a.s.l., 160 km, Servicio Meterológico Nacional Argentina; Matienzo: 64°58′S 60°04′W, , 25 m a.s.l., 70 km, Instituto Antártico Argentino - Glaciology Department

42 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula 7.5.2 Surface velocities Figure 7-3a (circles) shows the temporal trend (1993–2014) of the median ice flow velocity 500 m behind the glacier terminus. The surface velocity near the terminus increased from 0.65 m/d in 1993 at precollapse conditions to up to 8.8 m/d in 1999 after ice shelf disintegration. Subsequently, the median flow speed at the glacier tongue decreased to 1.5 m/d at the beginning of 2014. The very fast increase of the ice flow by a factor of about 5.5 within 1996 a few months to one year after the ice shelf disintegration is remarkable, culminating in a speedup of 13.5 times the precollapse surface velocities in 1999. Some short-term accelerations of 0.5–1 m/d were inferred at the glacier front in austral summers 2005, 2007, 2010, 2011, and 2012. Surface velocities between 2004 and 2008 show the highest variability. In the last five years of monitoring, observed flow speed at the terminus were still at least twice as high as under precollapse conditions. The surface velocity measured at the center of the flux gate on Edgeworth Glacier (black point, Fig 7-1) and its temporal variability is plotted in Fig. 7-3a (diamonds). A maximum velocity of 1.57 m/d was obtained at this location in September 2000. At the beginning of 2014, the velocity was even 10% lower than before the disintegration of Larsen-A Ice Shelf. 7.5.3 Surface elevation changes Figure 7-4 shows the elevation change (dh) between the ASTER DEM 2003 and the TDX DEM from March 21, 2014. The former trimline is also plotted, indicating the precollapse surface elevation. A considerable drawdown in the tributary glaciers 10 to 20 years after the ice shelf collapse is still visible amounting to more than 40 m during the observation period in some areas. The grounding line at Dinsmoor Glacier could be easily detected by the abrupt change in dh, while a more irregular and gentle change of dh was observed at the other two glaciers, making the GL estimation more difficult. Surface elevation profiles on Dinsmoor and Edgeworth Glacier (Fig. 7-4; Supplemental material, Fig. 13-5) show a change in slope at the supposed GL position, supporting our estimate. Subset areas on DBE were selected to investigate the temporal development of dh, considering the spatial and temporal limitations of the DEMs. Subset area S1 (light blue polygon, Fig. 7-4) covered the widest parts of DBE in the period 2003–2014 (103 km²; 17–1011 m a.s.l.). Subset area S2 (dark blue polygon, Fig. 7-4) is a subsection of S1, also covered by the ASTER DEM in 2001. Surface elevation changes were calculated with respect to the ASTER 2003 DEM. This DEM was chosen as a reference, because it is the earliest available DEM that covers the entire DBE system and was referenced to ICESat GLAS data. The median surface lowering on DBE is plotted in Fig. 7-3c. Between 1995 and 2003, a higher decrease was observed compared to the period afterwards. Between Oct. 2003 and Jan. 2014, we revealed a dh of 23.1±5.0 m on S1. We applied an exponential curve fit to extrapolate dh towards the precollapse situation. The curve fit matches with elevations derived from the trimline and the SPOT SPIRIT DEM as well as with dh measured on S2 between 2001 and 2003. An average surface lowering of ~130±15 m was determined between 1996 and 2014 at the glacier area covered by the DEM95. During the much denser TDX observation period 2011–2014, the average surface lowering rate amounted to 1.9 m/a. The detailed plot of the TDX data (Fig. 7-3c, inset) shows a pronounced seasonal signal with surface lowering in austral winters and elevation gain in austral summers. 43 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

Figure 7-4: Glacier surface elevation change between 2003 and 2014 at DBE. Black lines: Surface elevation profiles (see supplemental material, Fig. 13-5); red lines: Pre-collapse glacier trim line mapped on Landsat 5 image from March 1, 1986; purple dashed line: Estimated grounding line position in 2014; brown line: November 14, 2011 Operation IceBridge (OIB) Multichannel Coherent Radar Depth Sounder (MCoRDS) ice thickness profile on Edgeworth Glacier; dark and light blue polygons: Subset areas for surface elevation change analysis; background: Landsat LIMA Mosaic © USGS.

7.5.4 Ice mass flux The temporal variation of the ice mass flux (F; sum of all three glaciers) is shown in Fig. 7-5 for both bedrock elevation data sets. As with the ice-thickness data indicated, F based on the Huss and Farinotti (2014) bedrock revealed considerably higher values compared to F based on Bedmap2 data. A significant increase of F was observed after the ice shelf disintegration, like for the surface velocities. F based on Huss and Farinotti bedrock showed a strong increase from F = 0.77±0.12 Gt/a (mean 1993–1996) to 1.80±0.16 Gt/a in 2000 and a decrease to 0.73±0.07 Gt/a in early 2014. Hence, current values are slightly lower than under precollapse conditions.

The total ice flux (MF) since the breakup of Larsen-A Ice Shelf is estimated by specifying two possible peak function fit curves for F based on Huss and Farinotti bedrock. The maximum observed flux defines the peak of the lower bound solution, whereas the upper bound solution is defined by a strong increase of F in 1996 and the peak in 1998 (Fig. 7-5). The latter solution is supported by the temporal trend of the surface velocity at the glacier front. Between 1995 and

44 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

2014, MF resulted in −20.3 Gt and −25.0 Gt and the flux imbalances (MQ) in −6.4 Gt and −11.1 Gt, assuming that DBE was in equilibrium before 1995.

Figure 7-5: Temporal changes of ice mass flux through the Operation Icebridge profile on DBE using bedrock elevations from Huss and Farinotti (2014) and Bedmap2. Flux data based on Huss and Farinotti bedrock were interpolated to estimate mass loss MQ due to increased flux since 1995. Mean flux before 1995 is taken as balanced flux. The fit curves represent the inferred upper and lower bound estimates of MQ.

7.5.5 Mass balance The results for the estimates of the mass losses ΔM and ΔMslr of DBE between 1995 and 2014 are listed in Table 7-3. For the periods 1995–2003 and 2003–2014 the findings are listed in Tables 13-3 and 13-4, supplemental material. The fraction of Asl downstream of the flux gate was composed of the mass loss of −6.5±0.4 Gt above sea level and −5.6±0.4 Gt below sea level (1995–2014). In the period 2003–2014, the surface on the floating section (39.6 km²) lowered by 5.2±5.0 m, which corresponds to a submarine mass loss of −1.2±1.1 Gt. The average ice-thickness values of ti(max) =

723 m and ti(min) = 532 m as well as surface elevation of h95 = 108±10 m at the retreat area plus the

2014 mean freeboard height of 29.8 m of the glacier readvance area were used to estimate Af.

Table 7-3a shows that the highest contributions to ΔMg and ΔMIO resulted from Asl and Af, respectively, and that all findings depend strongly on the scenario of mass loss due to frontal retreat. The upper and lower bound approximations of the total mass flux described in Section

7.5.4 were used to calculate MQ. An accuracy of 1 Gt was estimated for MQ.

Applying upper and lower bound scenarios of Af and MQ to calculate ΔM of DBE since the breakup of Larsen-A Ice Shelf resulted in two estimates for the geodetic mass balance method and four results for the input/output method, ranging between −36±3.9 and −52.2±5.6 Gt (Table 7-3a). The contribution to SLR of all scenarios revealed results between 12.5±1.2 and 22.1±2.7 Gt. The results from the geodetic method are in the upper range or higher than mass losses followed from the input/output approach.

45 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula

Table 7-3: a) Total ice mass balance ΔM and b) contribution to sea level rise ΔMslr between 1995 and 2014 for different boundary conditions and methods. See Section 7.4.5 for detailed description of input parameters. Uncertainties are given with one σ.

7.6 Discussion 7.6.1 Velocity and elevation changes Our findings reveal the rapid reaction of the glacier system to the missing buttressing effect exerted by the former ice shelf and subsequent adaptation to the new situation. The results are consistent with observations by other authors. Rott et al. (2014) reported a surface velocity of 1.55 m/d at the center of the glacier front in 2013. Between 1996 and 2000, only one tracking result of 8.8±1.9 m/d (not plotted in Fig. 7-3a) was obtained at the terminus in Nov. 1999 during the ERS-Tandem

Mission. This result supports the upper bound estimation of MQ, even though no ice flux could be derived from it due to insufficient spatial coverage of the tracking results at the FG. Considering the error range and the location close to the glacier front, our maximum surface velocity is comparable to results published by Rott et al. (2002), giving values of up to 6 m/d 2 km upstream of the glacier front on a central flow line. They also reported flow velocities of 1.5 m/d 2 km upstream of the front and 2.9 m/d at the center of the front in 1995. Our velocity measurements for 1995 show a peak at the center of the front (width ~1.5 km) of up to 3.5 m/d, but it was averaged out in the median values along the front presented in Fig. 7-3a. We conclude that the acceleration and frontal retreat across the whole glacier front most likely started in late 1995 to early 1996. The delayed response of the glacier flow to the ice shelf collapse might have been induced by the inertia of the glacier system on the one hand, but also by a compact ice mélange having formed in the glacier embayment during winter 1995 on the other hand. It could have led to a short-term stabilization of the glacier front until the breakup of the ice mélange in austral summer 1996. This assumption is supported by visual inspection of the SAR intensity images from Oct. 1995 to March 1996. Calving events and changes in the buttressing force of the downstream ice mélange on the terminus probably caused the short-term retreats and accelerations at the glacier front observed since 2003 (see comparable observations for Jakobshavn Isbræ in Greenland by Amundson et al. (2010). Inspections of the SAR intensity images from austral summers show a preserved sea ice

46 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula cover in the fjord but with a reduced extent during speedup and calving events. The formation of cracks and polynyas in the ice mélange, induced by summer melt and foehn events, were probably most important. However, the retreat events did not occur every summer and have been generally smaller since 2007, with no significant retreat since 2012. This is supported by our own visual impression during survey summer flights from Marambio. The temperature records at Marambio and Matienzo stations (Fig. 7-3d) show a marked drop in the seven year average surface air temperature in 2007 by −1.9 °C (2000–2006: −9.4 °C, 2007–2013 −11.3 °C). The lower temperatures might have led to a thicker ice mélange, which was less frequent and easily broken up by melt and foehn events in summers. The observed surface lowering rate during the period 2003–2008 (dh/dt = −3.15 m/a) was 66% higher than that between 2011 and 2014. This can be addressed partly to a potential grounding line retreat. Our high values during the 2003–2008 period are comparable to, or lower than, those reported for the larger glacier systems located farther south such as Drygalski Glacier (−5 m/a 2004–2006) and Jorum Glacier (~−15 m/a, 2004–2006) (Shuman et al., 2011). The absolute elevation loss since 1995 of at least 130 m on the lower parts and the shape of adaptation is comparable to observations in Larsen-B catchments. Shuman et al. (2011) reported elevation change rates of −15 to −25 m/a for Hektoria and Green Glaciers immediately after the breakup of Larsen-B Ice Shelf (2001–2004). Thinning rates of 25 m/a were given for and for Jorum Glacier 60 m elevation loss (2002–2008) were observed. The latter is in the range of our approximation of DBE's elevation lowering after Larsen-A Ice Shelf disintegrated. Rott et al. (2014) reported a slightly higher mean dh/dt of −2.2 m/a for DBE in 2011–2013 than we observed, whereas Scambos et al. (2014) determined a mean dh/dt of −1.91 m/a (areas below 1000 m a.s.l., including Sobral Peninsula) for 2003–2008. The deviation from both studies is most likely caused by differences in spatial extent of the observed area and/or partly different data sets and processing approaches. Rott et al. (2014) assumed a floating area of 22 km² on Edgeworth Glacier; that is half of our estimate in 2014 including Dinsmoor Glacier and the frontal advance of 3.2 km² since 2013. The TDX time series analysis revealed a pronounced seasonal pattern, which is most probably caused by changes in the depth of the scattering phase centers. During summer, surface melt causes the presence of liquid water in the snow pack and following a reduction of the X-band SAR signal penetration depth to minimum. These radar penetration effects are on the order of 2–3 m. Even though this effect is smaller than the accuracies of our dh measurements, it could be minimized by comparing TDX DEMs from the same season or when comparing TDX DEMs with stereoscopic DEMs, using TDX data sets during melt conditions. In this study, we did not correct for penetration effects of the X-band SAR signal, since we assume almost zero penetration during summer melt situations and hence comparable conditions as by optical remote sensing. 7.6.2 Mass loss and Imbalance The methods and scenarios used to estimate ΔM and ΔMslr show large variabilities of up to 45% and 77%, respectively (period 1995–2014). Our findings from the geodetic method stretch across a smaller interval and are within the upper range or higher than input/output method results. The

47 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula approximation of Af, due to the unknown ice-thickness in the retreat area, is the major contribution slr to the variability of ΔM, whereas the influence of MQ on ΔM is most significant. At Dinsmoor Glacier, the GL estimate is quite reliable. Whereas at Edgeworth Glacier, the change in slope shifted slightly upstream between 2003 and 2014 (Supplemental material, Fig. 13-5) and the surface lowering continued, indicating a potential retreat of the GL. However, we did not consider in the mass balance computations that some sections upstream of our GL estimate on Edgeworth and Dinsmoor Glacier are presumably floating and that the GL probably retreated between 2003 and 2014. We also did not account for firn densification that might have contributed to a certain amount to the observed elevation change. However, we consider the influence of firn compaction small since at low elevations regular melt situations are known and field surveys revealed thin firn layers on the lower parts of the glaciers. The lower boundary surface mass balance scenario is quite unlikely, considering the high accumulation rates of 2350 kg/(m²a) observed at Laclavere Plateau (Fernandoy et al., 2012). We expect similar rates at the AP plateau near DBE and an accumulation decrease towards lower, lee- side elevations on the east side. Furthermore, we suppose that mass loss by surface melt might not be sufficient to reduce the mean catchment surface mass balance to 1080 kg/(m²a). The maximum accumulation rates of Fernandoy et al. (2012) might be too high for DBE, hence, we consider bsfc(max) to be the most plausible value for the surface mass balance. It is also close to the reported RACMO-2.1/ANT (Lenaerts et al., 2012) regional climate model results by Scambos et al.

(2014). To estimate MQ, we supposed that DBE was in equilibrium before 1995. This corresponds to a surface mass balance of bsfc5 = 1530 kg/(m²a) (Supplemental material, Section Error: Reference source not found-S1). It ranges between our approximated mass balance scenarios

(bsfc(min), bsfc(max)). From the wide variety of bsfc estimates and the discrepancies between the bedrock data sets, we conclude that bsfc and ice-thickness are the largest uncertainty factors of MQ. Overall this implies that improved and more reliable bsfc and ice-thickness data are most important to reduce the uncertainty of mass balance estimates. Although, we cannot give an absolute assessment of the validity of the different scenarios used, we will assess their plausibility in below. The upper bound value of MQ and the bedrock topography as measured by MCoRDS at the glacier retreat area are assumed to be the most reliable (scenario

5). The temporal trends of vs, Sr, and dh imply that the most reasonable value of MQ is the upper bound approximation. Both the temporal changes of the ice-thickness and the surface velocities show a similar pattern, although the data density in the most important time interval after the ice shelf disintegration is low. The two curve fittings do not cover the smaller variations in the flux and hence induce additional uncertainties. Furthermore, we assume for the fjord topography that the glacier tongue of DBE was well grounded in 1995. Taking those considerations into account slr scenario 5 with ΔMIO = −40.7±3.9 Gt and ΔMIO = 18.8±1.4 Gt in the period 1995–2014 (ΔMIO = slr −1.9±1.7 Gt and ΔMIO = 1.3±0.3 Gt in the period 2003–2014) provides the most plausible result for the input-output method. However, there are still uncertainties depending on the scenario of Af, slr bsfc, and MQ. For the geodetic mass balance method, ΔMg = −45.6±4.4 Gt and ΔMg = 22.1±2.7 Gt slr from scenario 1 are very probable in the period 1995–2014 (ΔMg = −3.6±2.2 Gt and ΔMg = 2.7±1.2 Gt, in the period 2003–2014). The results from the most likely scenarios, 1 and 5, overlap

48 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula at the 1-sigma level. For comparison with our ΔMslr estimates, Shepherd et al. (2012) gave a reconciled estimate of sea level contribution of 29±12 Gt/a (2000–2011) and 36±10 Gt/a (2005– 2010) for the entire Antarctic Peninsula. King et al. (2012) report −34±10 Gt/a (2002–2010) mass change for alone, whereas the mass loss on the western side of Graham Land seems to be quite small (Kunz et al., 2012). Considering the size of the study sites, our estimate of contribution to SLR for the period 2003–2014 is between the values given by King et al. (2012) and Shepherd et al. (2012). In this context, it has to be considered that Shepherd et al. (2012) reported predominately ungauged catchments for Larsen-A. Hence, the reliability of their values is further reduced. Other studies have inferred the mass balance of DBE for certain epochs, but not for the whole time span since Larsen-A Ice Shelf disintegrated. Rott et al. (2014) and Scambos et al. (2014) reported dM/dt = −0.523 Gt/a (2011–2013) and dV/dt = −0.87 km³/a (2003–2008, below 1000 m a.s.l.), respectively. Our measurements for the period 2003–2014 are lower than those, which is probably caused by the different spatial extend of the study areas (see also Section 7.6.1), whereas our mean annual mass loss over the entire period (1.9 Gt/a to 2.7 Gt/a) is significantly higher. The latter is in the range of the mass flux of large systems like Hektoria-Green (2.9 Gt/a, 2008) or Crane Glacier (2.9 Gt/a, 2008; 1.7 Gt/a, 2013) draining into former Larsen-B embayment six years after ice shelf breakup (Rott et al., 2011; Wuite et al., 2015). This supports the finding that the main ice loss occurred in the first years after the ice shelf breakup and decreased afterward. DBE's behavior is comparable to the response of Crane Glacier to the disintegration of Larsen-B Ice Shelf in 2002 (Rott et al., 2011; Scambos et al., 2004; Wuite et al., 2015). The flux imbalance estimate in early 2014 was −0.4% of the precollapse situation indicating that DBE has nearly adjusted to the new situation and that the mass flux is almost back to precollapse conditions. However, the continuous but relatively small surface lowering in recent years indicates that the ice flux is still not completely in balance with the bsfc and we conclude that bsfc has consequently decreased since 1995. Furthermore, the surface velocities at the front remained considerably higher than those under precollapse conditions. 7.7 Conclusions We showed a comprehensive analysis of the available SAR data time series with regard to surface velocities and frontal retreat as well as various optical and interferometric products for surface elevation change of the DBE glacier system. Our results accord with other studies conducted in the region, but our analysis covers a much longer time period starting at precollapse-ice-shelf conditions to subsequent adjustments until 2014, at high temporal resolution. A very steep increase in ice flow could be determined shortly after the ice shelf collapse in austral summer 1995. Since 2003, the previous continuous retreat has changed to an advance/retreat pattern most likely induced by calving and stabilization of the glacier front by the ice mélange in the fjord. Current velocities at the front are still considerably higher than those of precollapse flow, and ice-thickness has decreased by about 130 m since 1995. Our results show a rapid surface lowering after 1995, reaching lower elevation change rates of a few meters in recent years and a flux imbalance of −0.4% in 2014.

49 Study 1: Changes in ice dynamics, elevation and mass discharge of Dinsmoor- Bombardier-Edgeworth glacier system, Antarctic Peninsula The derived ice mass balances show significant variability depending on assumptions made and methods used. Most plausible is an overall mass loss (1995–2014) of −40.7±3.9 Gt and a contribution to SLR of 18.8±1.8 Gt, corresponding to 0.052±0.005 mm SLR equivalent. Our analysis and scenario considerations also reveal that major uncertainties are caused by a lack of sufficient accurate ice-thickness data. This relates to ice-thickness at the flux gate, but in particular also to ice-thickness in the retreated areas, leading to the largest uncertainties in the mass loss computations. Additionally, surface mass balance is still not well known yet. Climate models do not yet provide sufficient spatial resolution, and in-situ measurements do not exist for the DBE glacier system or surrounding catchments. Nevertheless, the dense time series analysis enables a better possibility to confine estimates of glacio-isostatic uplift at the northern AP caused by ice mass loss (Nield et al., 2014). Our results demonstrate that investigations considering only changes between specific years do not reveal the full picture of change and underlying processes. A comparison with results from time series of the Gravity Recovery and Climate Experiment (GRACE) satellite mission is much better facilitated by such denser time series. The presented details further improve our understanding of the reaction of tributary glaciers to ice shelf collapse.

Author contributions

T.S. analyzed all SAR, optical and elevation data. He performed the mass flux and imbalance computations. S.M. organized the field logistics, provided GIS data and supported the DEM analysis. V.H. processed the CryoSat-2 altimeter data. P.S. co-lead the study and supported the field surveys. M.B. initiated the project, coordinated the research and wrote the manuscript jointly with T.S.. All authors revised the manuscript.

Acknowledgements

M.B. and T.S. were funded under grant BR 2105/9-1 within the DFG Priority Program 'Antarktisforschung', and would like to thank the HGF Alliance “Remote Sensing of Earth System Dynamics”. S.M. was supported by different projects and programs funded by IAA-DNA. IAA/DNA and AWI kindly provided logistic support as well as airborne operations. The authors particularly acknowledge the support of Marambio station personnel, the helicopters and Twin Otter DHC-6 crews for their support during the field surveys. Access to satellite data was kindly provided by various space agencies, e.g. under ESA AO 4032, DLR TerraSAR-X Background Mission Antarctic Peninsula & Ice Shelves, TSX AO LAN0013, TSX AO BMC, TDX AO XTI_GLAC0264, ASF as well as NASA and USGS. Furthermore, this project has been supported by the European Commission th under the 7 Framework Programme through the Action – IMCONet (FP7 IRSES, action no. 319718). The authors would like to thank Matt King, Ted Scambos and 3 other anonymous reviewers for their constructive comments that helped improving this manuscript as well as the editor Derek Vance. Furthermore, we are grateful to Melanie Rankl, Saurabh Vijay and Björn Saß for helpful comments.

50 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series

8 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series

Seehaus Thorsten1*, Marinsek Sebastián2,3, Skvarca Pedro4, van Wessem Jan Melchior5, Reijmer Carleen H.5, Seco Luis José2, Braun Matthias1

1 Institute of Geography, Friedrich-Alexander-University Erlangen-Nuremberg, Wetterkreuz 15, D- 91058 Erlangen, Germany 2 Instituto Antártico Argentino, Balcarce 290, C1064AAF, Buenos Aires, Argentina 3 Universidad Tecnológica Nacional, Facultad Regional Buenos Aires, Medrano 951, C1179AAQ, Buenos Aires, Argentina 4 Glaciarium, Museo del Hielo Patagónico, El Calafate 9405, Prov. Santa Cruz, Argentina 5 Institute for Marine and Atmospheric Research Utrecht, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands Published in Frontiers in Earth Science (2016)

Abstract

The substantial retreat or disintegration of numerous ice shelves has been observed on the Antarctic Peninsula. The ice shelf in the Prince Gustav Channel has retreated gradually since the late 1980s and broke up in 1995. Tributary glaciers reacted with speed-up, surface lowering and increased ice discharge, consequently contributing to sea level rise. We present a detailed long- term study (1993-2014) of the dynamic response of Sjögren Inlet glaciers to the disintegration of the Prince Gustav Ice Shelf. We analyzed various remote sensing datasets to identify the reactions of the glaciers to the loss of the buttressing ice shelf. A strong increase in ice surface velocities was observed, with maximum flow speeds reaching 2.82±0.48 m d-1 in 2007 and 1.50±0.32 m d-1 in 2004 at Sjögren and Boydell glaciers respectively. Subsequently, the flow velocities decelerated, however in late 2014, we still measured approximately twice the values of our first observations in 1996. The Sjögren Inlet glaciers retreated 61.7±3.1 km² behind the former grounding line in 1996. For the glacier area below 1000 m a.s.l. and above the 2014 grounding (399 km²), a mean surface lowering of -68±10 m (-3.1 m a-1) was observed in the period 1993-2014. The lowering rate decreased to -2.2 m a-1 in the period 2012-2014. Based on the surface lowering rates, geodetic mass balances of the glaciers were derived for different time periods. A strongly negative mass change rate of -1.16±0.38 Gt a-1 was found for the area of all Sjögren Inlet glaciers (including the area above 1000 m a.s.l.) above the 2014 grounding line (559km2) for the earliest period (1993- 2001). Due to the dynamic adjustments of the glaciers to the new boundary conditions the rate changed to -0.54±0.13 Gt a-1 in the period 2012-2014, resulting in an average mass change rate of -0.84±0.18 Gt a-1 (1993-2014) for the same domain. Including the retreat of the ice front and grounding line, a total mass change of -37.5±8.2 Gt (-1.79±0.39 Gt a-1) and a contribution to sea level rise of 20.9±5.2 Gt (-0.99±0.25 Gt a-1) were computed for the period 1993-2014. Analysis of the ice flux revealed that available bedrock elevation estimates at Sjögren Inlet are too shallow and are the major uncertainty in ice flux computations. This temporally dense time series analysis of 51 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series Sjögren Inlet glaciers shows that the adjustments of tributary glaciers to ice shelf disintegration are still ongoing and provides detailed information of the changes in glacier dynamics. Keywords: Prince Gustav Channel, glacier mass balance, Antarctic Peninsula, sensitivity analysis, remote sensing, glacier change, contribution to sea level rise 8.1 Introduction On the northern Antarctic Peninsula substantial atmospheric changes have taken place in recent decades. A considerable increase in air temperature was reported, for example, by Skvarca et al. (1998) and Turner et al. (2005). Morris and Vaughan (2003) related the viability of ice shelves to the -9 degree Celsius annual isotherm, which shifted southward with changing climatic conditions. Scambos et al. (2003) linked the instability and recession of ice shelves to an increase in summer air temperatures and surface melt. Shepherd et al. (2003) also reported previous thinning, indicating substantial basalt melt. Numerous ice shelves (e.g. Larsen A/B, Larsen Inlet, Prince Gustav and Wordie) along the Antarctic Peninsula have significantly retreated or disintegrated since the late 1980s (Cook and Vaughan, 2010). Former ice shelf tributary glaciers reacted with accelerated ice flow and surface lowering to the loss of the buttressing ice shelf (e.g. De Angelis and Skvarca, 2003; Rack and Rott, 2004, 2003; Rignot et al., 2004; Rott et al., 2007, 2002; Wendt et al., 2010). Subsequently, significant glacier surface lowering was observed by Berthier et al. (2012) and Scambos et al. (2004) as a consequence of increased ice discharge. Rott et al. (2011) and Wuite et al. (2015) quantified the ice discharge of the Larsen B outlet glaciers using various remote sensing data. Along the Nordenskjöld Coast, Rott et al. (2014) derived a mass change rate of -4.2±0.4 Gt a-1 for the period 2011-2013 from bi-temporal TanDEM-X data. Based on photogrammetric digital elevation models (DEMs), a mass change rate of -24.9±7.8 Gt a -1 (2003- 2008) for the northern Antarctic Peninsula was found by Scambos et al. (2014). Ivins et al. (2011) reported a current ice mass change rate of -41.5±9 Gt a-1 from analysis of GRACE time series and bedrock uplift data. Shepherd et al. (2012) performed an integrated glacier mass balance compilation of the Antarctic and Greenland ice sheets based on modeling results and observation data. The authors revealed a mass change rate of -20±14 Gt a-1 (1992-2011) at the Antarctic Peninsula. The regional estimates show a significant mass loss at the Antarctic Peninsula, however, the individual imbalance results vary strongly and have uncertainties of up to 70%. Seehaus et al. (2015) performed a detailed time series analysis of various remote sensing datasets for the Dinsmoor-Bombardier-Edgeworth glacier system. The authors compiled the results in a comprehensive study of the effects of ice shelf disintegration for former Larsen-A tributary glaciers in the period 1993 to 2014. They identified major sources of uncertainties in mass loss computations. Furthermore, significant temporal variations of ice dynamics and mass loss were found, indicating that bi-temporal data analysis only partly reveals the dynamic response of tributary glaciers. The aim of this study is thus to analyze the dynamic adjustments of the Sjögren Inlet glaciers since the disintegration of the Prince Gustav Ice Shelf, as well as to examine the temporal change in imbalance and to quantify the ice mass loss for the whole study period 1993- 2014.

52 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series 8.2 Study site

Figure 8-1: Panels a and b: Location of Sjögren Inlet on the Antarctic Peninsula and retreat states of the Prince Gustav Ice Shelf. Map base: Landsat LIMA Mosaic © USGS and SCAR Antarctic Digital Database, version 6.0. Panel c: Surface velocity field and glacier front variation of Sjögren Inlet glaciers. Surface velocities were derived from TerraSAR-X/TanDEM-X acquisitions (December 11 and 22, 2012). Colored lines: Changes of glacier front position picked from SAR intensity images. Black dashed-dotted line: Grounding line position in 1996 (Rignot et. al., 2011). Blue triangles: Positions of surface velocity measurements. Black polygons: Glacier catchments from SCAR Antarctic Digital Database, version 6.0. Background: Landsat 8 image (band 4) from August 30, 2014 © USGS (for hypsometric curves of glacier catchments see Figure 8-3).

The Prince Gustav Ice Shelf was the most northerly ice shelf in Antarctica, covering an area of ~1600 km² in 1957 between James Ross Island and the Antarctic Peninsula (Figure 8-1,8-2B). It retreated gradually and broke up in 1995 (Cook and Vaughan, 2010; Cooper, 1997; Ferrigno et al., 2006; Skvarca et al., 1995) with the subsequent retreat, thinning and acceleration of former tributaries in the Röhss Bay on James Ross Island and Sjögren Inlet on the Antarctic Peninsula (Glasser et al., 2011; Rau et al., 2004; Rott et al., 2014). In 1993, the Prince Gustav Ice Shelf had retreated to the southern end of the mouth of Sjögren Inlet (ERS-1 SAR image 16 February, 1993; and in Rott et al., 1996 ERS-1 SAR image 26 August 1993). Subsequently, the ice front of Sjögren Inlet receded rapidly inwards of the bay. The retreat states between 1993 and 2014 are show in Figure 8-1. The catchment of Sjögren Inlet covered an area of about 560 km² in 2014 and consists of five tidewater glaciers, where Sjögren and Boydell glaciers dominate. The glacier catchment delineations are taken from the Antarctic Digital Database ADD6.0 (Cook et al., 2014) and the glaciers are named according to Davies et al. (2012). The GAP11 and GAP12 glaciers defined in Davies et al. (2012) are merged in the ADD6.0 delineations to jointly form the Sjögren Glacier. The same definition is used in this study. The lower regions with gentle surface slopes are separated

53 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series from the Antarctic Peninsula plateau by escarpments and ice falls. A strong west-east gradient in precipitation from the plateau to the coastal regions is caused by strong westerly winds reaching over the peninsula’s mountain range (van Wessem et al., 2015). The study site is often affected by foehn-type wind situations (Cape et al., 2015; Marshall et al., 2006; Orr et al., 2004), inducing higher surface air temperatures and increased solar radiation (due to cloud free conditions) compared to the west coast of the peninsula. High surface melt and the formation of melt ponds at lower elevations are linked to these events. 8.3 Data and Methods A broad multi-mission remote sensing database is used to study the adjustments of the Sjögren Inlet glaciers forced by the disintegration of the Prince Gustav Ice Shelf. The processing and accuracy analysis are briefly described in the following. Details of the work flow, error analysis and specification of the sensors and datasets are provided in Seehaus et al. (2015). 8.3.1 Flow velocity and glacier extent Surface velocity fields and glacier retreat are derived by means of Synthetic Aperture Radar (SAR) image time series (1993-2014). Intensity offset tracking (Strozzi et al., 2002) is applied to consecutive image pairs to obtain displacement fields for the respective observation intervals. The accuracies of the velocity measurements are the sum of the uncertainties of the image co- registration (measured on stable reference points like rock outcrops) and the intensity offset tracking process (depending on image resolution, time interval and oversampling, according to McNabb et al., 2012). Glacier fronts are manually mapped on multilooked SAR intensity images. GAMMA Remote Sensing software is used for processing the SAR data. 8.3.2 Surface elevation changes The Sjögren Inlet glaciers are part of a ”super test site” in the TanDEM-X mission plan. Consequently, bi-static interferometric SAR data have been regularly acquired by the TanDEM-X satellites at the study site since 2011. We interferometrically derived DEMs from the TanDEM-X Coregistered Single look Slant range Complex datasets (DLR-IMF, 2012). In order to reduce potential phase unwrapping errors a differential interferometric approach is applied. The ASTER AP DEM (Cook et al., 2012) is used as a reference DEM and in combination with the TanDEM-X data a differential interferogram is computed. Afterwards, the differential interferogram is filtered, phase unwrapped and transferred into an absolute DEM by adding the elevation from the reference DEM. Finally the TanDEM-X DEM is geocoded and orthorectified. In combination with DEMs (Table 8-1) from the “SPOT 5 stereoscopic survey of Polar Ice: Reference Images and Topographies” (SPIRIT) project (Korona et al., 2009) and Terra ASTER (Level 3 products processed by NASA Land Processes Distributed Active Archive Center, LP DAAC), changes in surface elevation are calculated by DEM differencing. Artifacts in the stereoscopic DEMs, caused by for example, sensor saturation, low image contrast or clouds, are manually masked out. The ASTER DEM in 2001, the SPIRIT DEM in 2006 and the TanDEM-X DEMs are vertically adjusted at sea level. Since stereoscopic DEMs are less reliable on water surfaces and TanDEM-X data strongly de-correlates on water, the sea ice covered area in Sjögren Inlet bay is used as a reference area (mean standard deviation of the vertical offsets on reference area is 2.02 m). The tidal changes of the sea level derived from the Circum-Antarctic Tidal Simulations model CATS2008a_opt (an updated version of 54 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series the barotropic inverse tide model described by Padman et al., 2002) reach up to 3.74 m at the study site. A tidal correction is therefore included in the vertical DEM referencing. All other ASTER DEMs are vertically referenced to ICESat Geoscience Laser Altimeter System data (GLA06 L1B Global Elevation Data Version 33, Zwally et al., 2012) acquired at a time interval of ±20 days relative to the DEM acquisition. Figure 8-2b shows the spatial extent of the DEMs used and the ICESat coverage.

Figure 8-2: Panel (A): Glacier surface elevation change between September 26, 2001 and December 22, 2014 at Sjögren Inlet. Pink dotted lines, Pre-collapse trimline mapped on a Landsat 5 image from March 1, 1986. Green lines, Surface elevations profiles along Sjögren and Boydell glaciers (see Figure 8-5). Blue lines, Flux gates at Sjögren and Boydell glaciers (cross sections see Figure 8-7). Orange polygons, Rock outcrops (stable ground) used to assess the accuracy of the DEM referencing. Black polygons, Glacier catchments from SCAR Antarctic Digital Database, version 6.0. Dashed black polygons, Manually masked out artifacts in ASTER DEM from September 26, 2001. Background, Landsat 8 image (band 4) from August 30, 2014 © USGS. Panel (B), Spatial extent of DEMs used and coverage of stereoscopic DEMs by ICESat GLAS data (same color as the DEM extents for the respective dates) acquired at a time interval of ±20 days relative to the DEM acquisition (date format: yyyy-mm-dd). Striped polygon, Areas below 1000 m a.s.l. Light blue polygon, Area of elevation change measurements (dh), Purple polygon, Sjögren Inlet catchment area.

The number of ICESat measurements used for vertical referencing, the respective laser periods and the measured vertical biases of all DEMs are listed in Table 8-1. The obtained offsets of the TanDEM-X DEMs vary between -35.44 and -11.84 m, due to different satellite orbit geometries and phase unwrapping start location. After vertical referencing, the vertical uncertainty of the DEMs is assessed on stable rock outcrops (Figure 8-2, orange polygons). Pre-collapse surface elevations are determined by mapping the former trimline on a Landsat 5 image from March 1, 1986 and its elevation extracted from the SPOT SPIRIT DEM in 2006. The stereoscopic DEMs are less reliable in the accumulation areas on the Antarctic Peninsula plateau, e.g. due to low image contrast and

55 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series sensor saturation. Moreover, the steep transition from the lower regions to the plateau affects the processing of the TanDEM-X DEMs by causing phase unwrapping errors (phase jumps). Consequently, analysis of elevation change (dh) is limited to regions ranging from the ice front position in 2014 up to the escarpments of the Antarctic Peninsula plateau (0-1000 m a.s.l., Figure 8-2b, striped polygon). Some TanDEM-X DEMs are affected by layover, shadowing and foreshortening close to the steep bluffs. These areas are masked out. Table 8-1: Digital elevation models (DEMs) used to study surface elevation changes. Date format: yyyy-mm-dd; n: number of ICESat measurements used for vertical referencing; LP: ICESat Laser Operations Period; Δh: vertical DEM bias derived at sea level (incl. tidal correction) or from ICESat data. σh: average vertical offset on rock outcrops (Figure 8-2) relative to ASTER DEM in 2001.

Date Sensor n LP Δh [m] σh [m] 2001-09-26 ASTER ---a --- 9.61 --- 2004-09-20 ASTER 25 3A -0.82 4.54 2004-10-22 ASTER 856 3A 1.79 5.16 2005-11-10 ASTER 501 3D 5.37 6.14 2006-01-07 SPOT ---a --- 4.63 0.90 2006-10-28 ASTER 223 3G 6.12 0.15 2011-06-14 -35.44 <-> <-> 2014-12- TanDEM-X ---a --- 2.65b -11.84 22

areferenced to sea level bmean value of all 30 TanDEM-X DEMs 8.3.3 Mass balance Geodetic mass balances, according to Fountain et al. (1997), are derived from elevation change rate (dh/dt) measurements integrated over the glacier area and multiplied by an average ice density. In this study an average ice density of 900 kg m-³ is applied. Thin firn layer and bare ice surfaces were observed at lower elevations during field surveys. Firn compaction is thus considered to be insignificant, as was also previously found by Scambos et al. (2014). The mass balance calculations based on elevation change observations are limited to regions upstream of the glacier front position in 2014, which is assumed to correspond to the grounding line (see Section 8.4.2). The ice mass loss due to frontal retreat is computed separately and discussed in Section 8.5.2. At the Antarctic Peninsula plateau, we applied an elevation change rate of 0.34 ± 0.15 m a-1 found by Scambos et al. (2014) at regions above 1000 m a.s.l., since our measurements are limited to areas below (see Section 8.3.2). In areas not covered by our dh/dt analysis, volume change is derived by hypsometric interpolation at 50 m intervals using the observed mean elevation change rates for each altitude band. Figure 8-3 shows the hypsometric curves of Sjögren Inlet glaciers based on the ASTER AP DEM. The geodetic mass balance of all Sjögren Inlet glaciers combined and Sjögren and Boydell glaciers separately are calculated for five periods (four successive periods and the entire study period). To evaluate the glacier mass balance and its temporal changes, the imbalance ratio is calculated for different time steps. According to Scambos et al. (2014), we calculated the imbalance ratio (I) by dividing the geodetic mass balance by the climatic mass balance (bclim; surface mass balance

56 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series including internal accumulation according to Cogley et al., 2011). It indicates the level of imbalance of a glacier system.. Climatic mass balance is analyzed from the regional atmospheric climate model RACMO2.3 at a horizontal resolution of 27 km (van Wessem et al., 2014) and of 5.5 km (van Wessem et al., 2016). Both model versions cover the period 1979-2014 and are forced at the lateral boundaries by ERA-Interim reanalysis data (Dee et al., 2011). The 27 km grid size model version covers the complete Antarctic ice sheet, while the high resolution version only covers the Antarctic Peninsula, leading to slight but insignificant differences in the boundary forcing (van Wessem et al., 2016). The climatic mass balance data is extracted for the individual catchments, considering a weighting of partly covered model grid cells.

Figure 8-3: Hypsometric curves of Sjögren-Inlet glacier catchments based on the digital elevation model from Cook et al. (2012). Note the different scaling of the y-axes.

8.3.4 Ice mass flux The ice mass fluxes of Sjögren and Boydell glaciers are computed at the flux gates close to the estimated grounding line position in 2014 (Figure 8-2 and Section 8.4.2). Ice thickness at the flux gates is calculated by utilizing the Huss and Farinotti (2014) bedrock map of the Antarctic Peninsula and the TanDEM-X DEM from December 22nd, 2014. The bedrock map is based on ice dynamic modeling. Changes in ice thickness due to surface lowering are considered in the ice flux computations. Surface velocities and bclim (from RACMO2.1, Lenaerts et al., 2012) were used as input as well as various ice thickness measurements for calibration and validation. The precision of the ice thickness at the flux gates is estimated using the accuracy map of the bedrock dataset. 8.4 Results 8.4.1 Flow velocity and glacier extent Due to the continuous retreat of the Prince Gustav Ice Shelf, the Sjögren Inlet was no longer buttressed by the ice shelf in early 1993. Subsequently, the ice front retreated rapidly into the Sjögren Inlet bay until it reached an almost stable position in 2007. The area change is calculated relative to the grounding line position in 2006 from Rignot et al. (2011) (black dashed-dotted line

57 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series Figure 8-1). Its temporal course is plotted in Figure 8-4. The maximum retreat of -61.7±3.1 km² is mapped on the TanDEM-X SAR image from April 6, 2014, which correlates to the loss of grounded ice (see Section 8.4.2).

Figure 8-4: Variations of surface velocity and frontal retreat of Sjögren Inlet glaciers between 1995 and 2014. Panels a and b: Surface velocities derived by SAR intensity tracking at the central flow line close to the front position in 2014 at Sjögren and Boydell glaciers (See Figure 8-1, blue triangles). Note the different scale of the y-axes. Panel c: Glacier area changes due to frontal retreat relative to grounding line position in 1996.

The surface velocities of Sjögren and Boydell glaciers are measured at the central flow line position close to the ice front in 2014 (Figure8-1, blue triangle). Within a radius of 200 m at both locations, the median ice velocities are extracted from the offset tracking results. Figure 8-4 shows the temporal changes (1995-2014) in surface velocities of both glaciers. The earliest tracking results reveal 0.79±0.14 m d-1 for Sjögren Glacier in February 1996 and 0.38±0.08 m d-1 for Boydell Glacier in December 1995. Flow velocities of both glaciers increased strongly afterwards, reaching the highest observed values of 2.82±0.48 m d-1 for Sjögren Glacier in December 2007 and of 1.55±0.32 m d-1 for Boydell Glacier in November 2004. Afterwards, the flow velocities decreased

58 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series towards 1.44±0.03 m d-1 and 0.97±0.07 m d-1 at Sjögren and Boydell glaciers in December 2014, respectively. In the austral summers 2002/03, 2004/05, 2006/07, 2007/08, 2010/11, 2011/12 and 2013/14 some short-term speedups are detected at Sjögren Glacier, which mostly correlate with fluctuations in the calving front position. Before 2002 and in austral summers 2003/04, 2005/06, 2008/09, 2009/10 and 2012/13 the data density is insufficient to evaluate such short term changes. 8.4.2 Surface elevation changes Figure 8-2 shows the elevation differences between the ASTER DEM in 2001 and the latest TanDEM-X DEM from December 22, 2014. The surface lowering amounts to more than 100 m close to the calving front (2014) of Boydell Glacier and the mean measured lowering of all Sjögren Inlet glaciers is -34.1±5.0 m during this period. We assume the recent grounding line position of all Sjögren Inlet glaciers to be close to the glacier front. No elevation change pattern is detected, which would indicate a grounding line position more upstream of the glacier. This assumption is supported by an observed increase in the slope a few hundred meters behind the calving front. Elevation profiles along Sjögren and Boydell glaciers (Figure 8-2, green lines) are presented in Figure 8-5.

Figure 8-5: Ice surface elevation profiles at Sjögren and Boydell glaciers (see Figure 8-2). Breaks in slope on the lower part of the glaciers indicate the respective grounding line position. Date format: yyyy-mm-dd.

59 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series

Figure 8-6: Mean measured surface elevation changes of Sjögren Inlet glaciers at areas below 1000 m a.s.l. relative to the ASTER DEM in 2001. Note the different scale of the y-axes for the lower graphs.

Changes in surface elevation are calculated relative to the earliest available cloudless stereoscopic

ASTER acquisition on September 26, 2001. The vertical offsets σh of the DEMs relative to the ASTER DEM in 2001, calculated on rock outcrops, are listed in Table 8-1. DEMs are less reliable on steeper slopes (Toutin, 2002), which are typical for rock outcrops and nunataks. We thus used a maximum error of 5 m for our elevation change measurements, albeit the σh values of two ASTER DEMs are slightly higher. An inaccuracy of 10 m is estimated only for dh measurements derived from the DEM based on the trimline altitude. Figure 8-6 shows the mean measured surface lowering trend of Sjögren Inlet glaciers in the period 1993-2014. The ASTER DEMs in September 2004 and October 2006 are partly covered by clouds. The ASTER DEM in 2005 does not cover the whole study site, and therefore, not every ASTER DEM is used for each glacier to study elevation changes. It is supposed that the dynamic response to the disintegration of the Prince Gustav Ice Shelf started in early 1993, when the ice shelf had retreated behind the mouth of Sjögren Inlet bay. The derived pre-collapse surface elevation is thus taken as the starting point of the time series. At the GAP09, GAP10 and GAP14 glaciers, the pre-collapse surface elevation information only partly covers the glaciers. Therefore, the dh information of the covered area is interpolated by using the hypsometry (Figure 8-3) in order to calculate the average elevation change of all Sjögren Inlet glaciers between 1993 and 2001 (Figure 8-6). All glaciers show a clear surface lowering trend with a decrease in surface lowering rate in more recent years. The dense TanDEM-X time series (2011- 2014) reveals seasonal fluctuation in austral summers 2011/12 and 2013/14. Higher surface elevations (~3-5 m) are found between late November and March. In austral summer 2012/13, the data density is too low to evaluate seasonal changes.

60 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series 8.4.3 Mass balance

The computed geodetic mass balance above the 2014 grounding line, ΔMT of the total glacier area and mass changes of the regions below 1000 m a.s.l. (ΔML),of all Sjögren Inlet glaciers combined and Sjögren and Boydell glaciers separately are listed in Table 8-2. Table 8-2: Mass balance of all Sjögren-Inlet glaciers (A) and individually for Sjögren (B) and Boydell (C) glaciers for the area upstream of the grounding line in 2014 for different observation periods. ΔML: Mass balance of the glacier area AL below 1000 m a.s.l.; ΔMT: Mass balance over the area AT including the entire altitude range; bclim: climatic mass balance from RACMO 2.3 (5.5 km grid size) over AT; I: Imbalance ratio; Adh: Area below 1000 m a.s.l where elevation change observations are available

A) Sjögren Inlet Glaciers: AL = 399 km²; AT= 559 km²; Adh = 350 km² ΔM ΔM b B Period L T clim clim I [Gt a-1] [Gt a-1] [kg m-² a-1] [Gt a-1] Jan. 1993 - Sep. 2001 -1.21±0.36 -1.16±0.38 522 0.27 -4.3±1.4 Sep. 2001 - Jan. 2006 -1.00±0.37 -0.95±0.39 526 0.27 -3.5±1.4 Jan. 2006 - Jan. 2012 -0.49±0.26 -0.44±0.28 625 0.32 -1.4±0.9 Jan. 2012 - Dec. 2014 -0.59±0.11 -0.54±0.13 635 0.33 -1.6±0.4 Jan. 1993 - Dec. 2014 -0.89±0.16 -0.84±0.18 564 0.29 -2.9±0.6

B) Sjögren Glacier: AL = 196 km²; AT = 282 km²; Adh = 173 km² ΔM ΔM b B Period L T clim clim I [Gt a-1] [Gt a-1] [kg m-² a-1] [Gt a-1] Jan. 1993 - Sep. 2001 -0.84±0.18 -0.81±0.19 610 0.17 -4.7±1.1 Sep. 2001 - Jan. 2006 -0.59±0.18 -0.56±0.19 608 0.17 -3.3±1.1 Jan. 2006 - Jan. 2012 -0.36±0.13 -0.33±0.14 714 0.20 -1.7±0.7 Jan. 2012 - Dec. 2014 -0.34±0.05 -0.31±0.06 723 0.20 -1.5±0.3 Jan. 1993 - Dec. 2014 -0.59±0.07 -0.57±0.08 651 0.18 -3.1±0.5

C) Boydell Glacier: AL = 36 km²; AT = 80 km²; Adh = 29 km² ΔM ΔM b B Period L T clim clim I [Gt a-1] [Gt a-1] [kg m-² a-1] [Gt a-1] Jan. 1993 - Sep. 2001 -0.09±0.03 -0.08±0.04 722 0.058 -1.4±0.6 Sep. 2001 - Jan. 2006 -0.16±0.03 -0.15±0.04 700 0.056 -2.6±0.7 Jan. 2006 - Jan. 2012 -0.05±0.02 -0.03±0.03 799 0.064 -0.5±0.4 Jan. 2012 - Dec. 2014 -0.05±0.01 -0.04±0.01 798 0.064 -0.6±0.2 Jan. 1993 - Dec. 2014 -0.09±0.02 -0.07±0.02 747 0.060 -1.2±0.3

For the period 2012-2014, an error for the elevation change rate of 0.34 m a -1 is applied. It is derived by differencing the corresponding TanDEM-X DEMs on stable areas and is similar to findings by Rott et al. (2014). In order to calculate the imbalance ratio, the climatic mass balance from both RACMO2.3 model versions is extracted considering 4 and 31 grid cells for the 27 km and the 5.5 km model version, respectively. The high resolution model version reveals bclim values approximately three times lower than for the 27 km model version. Due to the strong topographic gradients and spatial variability of precipitation at the study region, the results from the high resolution RACMO2.3 model version are used in the further analysis (see also Discussions;

Section 8.5.2). The obtained mean bclim values and glacier wide climatic mass balances Bclim as well as the imbalance ratios are also listed in Table 8-2 for the respective time intervals and 61 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series catchments. The bclim shows a slight increase in recent years and the ΔMT is significantly less negative. Hence, the negative imbalance ratio decreased by a factor of about three and two compared to imbalance ratios of -4.7±1.1 and -1.4±0.6 at Sjögren and Boydell glaciers in the period straight after the collapse of the Prince Gustav Ice Shelf (negative imbalance ratios indicate the level of down wasting or excessive ice flux). The most recent imbalance ratio estimation of all Sjögren Inlet glaciers is approximately three times smaller than in the period 1993-2001, but still negative. 8.4.4 Ice mass flux The profiles of bed topography, surface elevation (in December 2014) and ice thickness at the flux gate of Sjögren and Boydell glaciers are shown in Figure 8-7. The ice thickness is corrected for the linear trend of -3.6 m a-1 (R²=0.94) determined at regions below 200 m a.s.l. Figure 8-8 shows the temporal trend of the ice flux of Sjögren and Boydell

Figure 8-7: Ice surface and bedrock elevation at the flux gates of Sjögren and Boydell glaciers (see Figure 8-2). Blue line: Bedrock elevation from Huss and Farinotti (2014). Green line: Surface elevation from TanDEM-X DEM in December 2014. Pink line: Ice thickness = difference between surface and bedrock elevation. glaciers. The horizontal lines indicate the mean bclim extracted from the RACMO2.3 model results in the study period. At the beginning of the study period, the mass flux of both glaciers is comparable

62 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series to bclim. After the disintegration of the Prince Gustav Ice Shelf it increased significantly, up to fourfold, with a subsequent decrease. In the last observation period, it was about two times higher than under pre-collapse conditions. A peak curve fitting is applied to the flux time series to estimate the total mass discharge since the beginning of the study period. At Sjögren and Boydell glaciers it amounts to 13.80 Gt and 2.77 Gt of ice, respectively.

Figure 8-8: Temporal variation of ice mass flux at flux gates at Sjögren and Boydell glaciers. Blue dashed lines: Mean climatic mass balance of respective glacier derived from RACMO2.3. Red dashed-dotted lines: Curve fitting of ice flux (peak function).

8.5 Discussion 8.5.1 Ice dynamics Since the Prince Gustav Ice Shelf retreated behind the Sjögren Inlet bay in 1993, the ice dynamics of the tributary glaciers have shown substantial changes. In the 1990s, the glaciers retreated and accelerated rapidly, with a subsequent decrease in flow velocity and a stabilization of the ice front positions in recent years. The obtained flow velocities and ice flux of Sjögren and Boydell glaciers were about 2-3 times higher in 2014 than in late 1995 and early 1996, indicating that ice mass loss

63 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series still persists. Unfortunately, no surface velocity information can be derived from SAR data before 1995 due to inadequate satellite image coverage. Similar dynamic adjustments of tributary glaciers to ice shelf retreat and disintegration were reported for other study sites (e.g. Rignot, 2005; Rignot et al., 2004; Rott et al., 2011; Wuite et al., 2015). Our observed surface velocities at Sjögren Glacier are consistent with observations made by De Angelis and Skvarca (2003) and Rott et al. (2014). Some small deviations are most likely caused by differences in the location of the measurements. Davies et al. (2012) reported higher rates of glacier length changes at the Prince Gustav Ice Shelf tributary glaciers and a faster recession of the Sjögren Inlet glaciers between 1988 and 2001, compared to the period 2001- 2009. Their findings are in accordance with ours, however the results presented in this study have a much higher temporal resolution. Maximum flow speeds are observed at Sjögren and Boydell glaciers 11 and 14 years after the retreat of the Prince Gustav Ice Shelf, respectively. The glacier front also stabilized its position after ~11 years of nearly continuous retreat. At Dinsmoor- Bombardier-Edgeworth glacier system, Seehaus et al. (2015) found the maximum velocity and a stabilization of the ice front position about 5-6 years after the disintegration Larsen-A Ice Shelf. The authors also observed a rapid increase in flow velocity shortly after ice shelf breakup. In contrast to Dinsmoor-Bombardier-Edgeworth glacier system, Sjögren and Boydell glaciers are located more inwards in the bay. The ice front at Sjögren Inlet was also about 10 km in 1993, and 3 km in 1996, away from the groundling line position in 1996. At Dinsmoor-Bombardier-Edgeworth glaciers system the ice front was close to the grounding line after disintegration of the Larsen-A Ice Shelf. Unfortunately, no information of the grounding line position in 1993 is available at Sjögren Inlet. We conclude that Sjögren and Boydell glaciers reacted more slowly to the loss of the ice shelf compared to Dinsmoor-Bombardier-Edgeworth glacier system, because of the different geometry of the glacier bays. Bedrock elevation data at the retreat areas would help to better interpret the different responses of the glaciers to nearly similar perturbations, but no reliable bathymetry data is available for either site. Since 2002, short term accelerations and retreats of the glacier front have been detected. They occur during austral summers. We have two hypotheses to explain these events. On the one hand side, the accelerations and retreats could be caused by calving events and a weakening of the buttressing effect of the ice mélange in the fjord due to summer melt. Visual inspection of the SAR imagery reveals the formation of cracks and a fragmentation of the ice mélange in Sjögren Inlet bay as well as areas with open water in austral summers 2002/03, 2004/05, 2006/07, 2012/13. In austral summers 2007/08, 2010/11 and 2013/14 the ice mélange was not fragmented. Lower amplitudes of the accelerations are observed in these summers. This finding supports our hypothesis, and similar conclusions were obtained by Seehaus et al. (2015) for Dinsmoor- Bombardier-Edgeworth glacier system as well as by Amundson et al. (2010) and Moon et al. (2015) in Greenland. On the other hand, these events could be connected to melt water runoff availability, as observed by Moon et al. (2014) in Greenland. During early summer, an inefficient or closed subglacial drainage system may cause lubrication of the entire glacier bed, and thus a reduction of the basal friction with a consequent glacier speedup. If high runoff is available in late summer, the subglacial drainage system is well developed, forcing a channelized melt water

64 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series discharge and a decline in glacier velocities as observed in 2005, 2007, 2008, 2010 and 2012 at Sjögren Glacier. Cape et al. (2015) analyzed meteorological records from the Argentine station Matienzo (~90 south of Sjögren Inlet) in the period 1962-2010. The authors found high numbers of foehn days in austral summers 2004/05, 2005/06, 2006/07 and 2007/08. Unfortunately, there are no observations available directly at Sjögren Inlet, but their findings support our hypotheses that high summer melt caused the acceleration and retreat events. A continuous trend in ice surface elevation change was measured between 1993 and 2014 (-4.0 m a-1 at Sjögren Glacier, -3.6 m a-1 at Boydell Glacier and -3.1 m a-1 for all Sjögren Inlet glaciers). The trend decreased in recent years, but still amounts to -2.51 m a-1 at Sjögren Glacier, and -2.17 m a-1 at Boydell for the period 2012-2014. Between austral winters 2011 and 2013, we observed a surface lowering rate of -1.90 m a-1 for all Sjögren Inlet glaciers, which is comparable to the findings of Rott et al. (2014) of -1.74 m a-1 for the similar period and observation area (the authors kindly provided their dh/dt grid). Scambos et al. (2014) found a value of -1.64 m a-1 for the period 2001-2006, which is significantly smaller than our observation of -3.45 m a-1. This divergence is most likely caused by the differences in the processing techniques and spatial extent of the observation area. The observation area of Scambos et al. (2014) also covers areas north-west of Sjögren Inlet and parts of Longing Peninsula, where relatively small elevation change values were observed. Our observed lowering rates are also comparable to, or smaller than, values reported for glacier systems in the Larsen-A/B region farther south. High surface lowering rates of -10 to -25 m a-1 at Hektoria, Green, Crane and Jorum glacier were found by Shuman et al. (2011) in the period 2001-2004. Seehaus et al. (2015) reported high surface lowering rates at Dinsmoor-Bombardier- Edgeworth glacier system immediately after the disintegration of the Larsen-A Ice Shelf in 1995 with subsequent smaller rates until 2014. This behavior is similar to that noted in our observations at Sjögren Inlet glaciers and indicates the dynamic adjustment of the tributary glaciers to the new boundary conditions. The surface lowering rate at Dinsmoor-Bombardier-Edgeworth glacier system was about 2 times higher than at Sjögren Inlet glaciers in the first years after ice shelf breakup and is nearly the same in the period 2012-2014. The total surface lowering at Dinsmoor-Bombardier- Edgeworth glacier system is also significantly higher. These findings are in accordance with the more rapid and intense acceleration of the glacier system after ice shelf breakup, compared to Sjögren Inlet glaciers (see above). Changes in the X-band SAR signal penetration depth most probably induced the seasonal changes in dh observed in the dense TanDEM-X data time series. Foehn wind events and summer melt cause a liquid water content in firn and on the snow surface. The scattering phase center of the SAR signal thus shifted close to the surface and penetration depth is minimal. By comparing TanDEM-X DEMs from the same season and comparing the stereoscopic DEMs with summer TanDEM-X DEMs (assuming nearly no penetration during summer melt conditions), we minimize the effect of SAR signal penetration on elevation change analysis (Seehaus et al., 2015). Based on the SPOT SPIRIT DEM in 2006, Davies et al. (2012) classified the glacier tongues at Sjögren Inlet, except for Boydell Glacier, as “partly floating” and “grounded”, concordant with our grounding line estimate. The tongue of Boydell Glacier was classified as “floating”, which agrees partly with our findings. The break in slope, that indicates the grounding line position, was

65 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series approximately 2 km upstream the glacier front in 2006 (Figure 8-5, lower panel, red line). Afterwards, the glacier front receded, whereas the break in slope moved backwards just slightly. In consequence, the grounding line position is currently close to the glacier front. Our own visual observations of the glacier tongue during survey flights support this assumption. 8.5.2 Mass balance The changes in ice dynamics after the disintegration of the Prince Gustav Ice Shelf caused a significant mass loss of its tributary glaciers. Between 1993 and 2014, 18.5±4.0 Gt of ice upstream of the current grounding line (2014) were lost by dynamic readjustments. The mass gain in the plateau regions is very small, compared to relative large mass change ΔML observed in the lower regions, and just slightly affects the glacier wide mass balance ΔMT. The mass balance calculations reveal high mass losses after the retreat of the ice shelf, with a successive decrease. At Dinsmoor-Bombardier-Edgeworth glacier system, Seehaus et al. (2015) observed a similar temporal evolution of the mass loss at regions below 1000 m a.s.l. (observed area: 610 km²) as a consequence of ice shelf disintegration. In the period 2011-2013, Rott et al. (2014) found a mass change rate for Sjögren Inlet glaciers of -0.364 Gt a-1 (≙ -1.91 Mt km-2 a-1 observed area: 190.2 km²) at regions below the Antarctic Peninsula plateau, which is comparable to our observations in the period 2012-2014 considering the size of the observation area (Table 8-2). Scambos et al. (2014) reported a volume change of -1.40 km³ a-1 ≙ -1.26 Gt a-1 (≙ -1.48 Mt km-2 a-1 observed area: 852.8 km²) in regions below 1000 m a.s.l. for the period 2001-2006. As mentioned above, deviations from our results are most probably due to differences in the observed area. All our observed ice mass loss, derived using the geodetic mass balance method, contributed to sea level rise, because it was measured on areas upstream of the grounding line position in 2014, however, the total contribution to sea level rise is certainly higher, since the grounding line has retreated significantly since 1996 (Section 8.4.2). The thinning of the glacier caused the retreat of the grounding line. A glacier tongue starts to float when the ice-thickness decreases below a threshold defined by the hydrostatic equilibrium, and depends on the water depth (bedrock elevation). Consequently, only the ice mass loss of grounded ice contributes to sea level rise (see also Seehaus et al., 2014, Figure 6.2). Quantifying the ice mass loss caused by the retreat of the ice front and grounding line is difficult, since no reliable bathymetry data is available in the fjord (see also last paragraph of this section). The bedrock datasets from Huss and Farinotti (2014) and Bedmap2 (Fretwell et al., 2013) differ strongly in the retreat area, presumably caused by differences in the applied modeling and interpolation approaches and boundary effects at the margins of the model domains. We assume that all ice was grounded upstream of the grounding line in 1996 and estimate a mean fjord depth of -260±66 m based on the average surface elevation (derived from trimline altitude analysis) at the former grounding line position in 1996 and the hydrostatic equilibrium assumption. Our estimate is larger than the average bedrock depth of -203 m and -193 m at the flux gates of Sjögren and Boydell glaciers (Figure 8-7), respectively, based on the bedrock grid from Huss and Farinotti (2014). In combination with the pre-collapse surface elevations at the retreat area, a total mass change of -19.0±4.2 Gt of previously grounded ice is found due to frontal retreat in the observation period. Just a small fraction of 2.4±1.2 Gt contributed to sea level rise, because the volume below sea level of the lost ice has to be taken into account.

66 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series Hence, the total mass change of -37.5±8.2 Gt and contribution to sea level rise of 20.9±5.2 Gt are comparable to values reported by Seehaus et al. (2015) for Dinsmoor-Bombardier-Edgeworth glacier system, even though larger surface lowering rates were found at the glaciers system. The size of the glacier catchments is comparable (~610 km²), but a larger fraction of Dinsmoor- Bombardier-Edgeworth glacier system covers the plateau, which is dynamically decoupled from the lower sections by escarpments and ice falls. The temporal trend of the imbalance ratio clearly indicates a significant decrease in the glacier draw down. The imbalance ratio depends strongly on the applied climatic mass balance, and therefore, good quality bclim data is necessary. Due to the steep topography of the Antarctic Peninsula and pronounced upwind-downwind effects the climatic mass balance shows a strong spatial variability (Turner, 2002). Literature values from Fernandoy et al. (2012) (Laclavere Plateau - -1 ~90 km northeast of Sjögren Inlet, bclim = 2350 kg m ² a ), Rott et al. (2011) (Crane Glacier ~200 - -1 km southwest of Sjögren Inlet, bclim = 1087±122 kg m ² a ) and Seehaus et al. (2015) (Dinsmoor- - -1 Bombardier-Edgeworth glacier system ~40 km south of Sjögren Inlet, bclim = 1070 - 1720 kg m ² a ) also indicate the great spatial variability along the Antarctic Peninsula. Consequently, comparison of the model results at Sjögren Inlet with the literature values from other locations is difficult. The main glaciers in the Sjögren Inlet, Sjögren and Boydell glaciers, have large, high-elevation accumulation areas at the plateau. Downstream of the easterly margins of the mountain ridge, the elevation drops strongly within several hundred meters. At Sjögren Inlet catchment, the mean climatic mass balance (1979-2014) of the RACMO2.3 model amounts to 938 kg m-² a-1 at regions above 1000 m a.s.l. and 418 kg m-² a-1 below 1000 m a.s.l. On these small scales the results are very sensitive to the horizontal resolution of the model. Even at the high 5.5 km horizontal resolution of RACMO2.3, not all topographic features are resolved, affecting precipitation and foehn winds (van Wessem et al., 2016) and hence bclim. Despite this, the 5.5 km grid size RACMO2.3 model version is currently the most suitable model for our study region. We assume that the Sjögren Inlet glaciers were in equilibrium before the breakup of the Prince Gustav Ice Shelf. The ice mass loss based on the curve fittings of the ice flux (Figure 8-8, red dotted dashed lines) is thus determined by subtracting bclim from the ice discharge. For the period 1993-2014, mass change rates for Sjögren and Boydell glaciers of -0.45 Gt a-1 and -0.067 Gt a-1, respectively, are found. For both glaciers the mass change rate obtained from the ice flux is lower than the results obtained by the geodetic mass balance method (Sjögren: -0.57±0.10 Gt a-1; Boydell: -0.07±0.02 Gt a-1). The largest source of uncertainty in the computed ice mass flux is the bed topography. Surface velocities and the surface elevation at the flux gate show only small errors, however, no ice thickness measurements are available at Sjögren and Boydell glaciers, which could have been used to constrain the bedrock modeling and to reduce its uncertainty at the study site. Considering the difference in mass change rates obtained from dh measurements and ice mass flux, we conclude that the ice thickness at the flux gates is underestimated by about 30%. 8.6 Conclusions This study presents a detailed long term analysis (1993-2014) of the response of tributary glaciers in the Sjögren Inlet to the retreat and disintegration of the Prince Gustav Ice Shelf. In early 1993, the Sjögren Inlet bay was no longer blocked by the ice shelf. As a consequence of the missing

67 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series buttressing by the ice shelf, the glaciers receded by -9.1 km² a-1 inwards of the Sjögren Inlet bay and the tributary glacier started to accelerate. During the last decade, the flow velocities slowed down, but are still higher than our earliest measurements in 1995/96. In contrast to Dinsmoor- Bombardier-Edgeworth glaciers system (Seehaus et al., 2015), the observed acceleration and surface lowering is less rapid and intense, which is most likely caused by the different settings of the glacier bay. In the period 1993-2001, the observed average mass loss rate is two times higher than in the time interval 2012-14. The continuous decrease of mass loss leads to a reduction in the negative imbalance in recent years, with imbalance ratios ranging between -0.6 and -1.6 in the period 2012-2014. Overall, a total ice mass change of -37.5±8.2 Gt and a contribution to sea level rise of 20.9±5.2 Gt is found for all Sjögren Inlet glaciers in the study period 1993-2014. The temporally detailed analysis of glacier mass balances can be used to better delimitate and refine the glacio-isostatic uplift, found to be significant in the northern Antarctic Peninsula (Nield et al., 2014), since bi-temporal investigations do not fully reflect the ongoing dynamic changes. Our observations point out that major uncertainties in mass balance calculations are caused by the imprecise bed topography data at the flux gates as well as at the glacier retreat area. The extreme elevation gradients and the high spatial variability of atmospheric parameters also suggest that the 5.5 km grid size RACMO2.3 model version currently provides the most reliable climatic mass balance database for the Antarctic Peninsula, where field measurements are still very sparse. The nearly stable glacier front positions since 2007, the deceleration of the ice flow and the decrease in ice mass loss indicate that the dynamic response of Sjögren Inlet glaciers to ice shelf disintegration is slowly declining, however, the effects of the disintegration of the ice shelves on the dynamics of tributary glaciers can be still seen more than 20 years afterwards, as shown in this study. The comparison of the results of this study at Sjögren Inlet, with the analysis of the Dinsmoor-Bombardier-Edgeworth glacier system by Seehaus et al. (2015), reveals clear differences in the reaction and adjustments of tributary glaciers to ice shelf disintegration, most likely controlled by local settings such as bedrock elevation and fjord geometry. The temporally detailed analysis of ice dynamics performed in both studies provides a basis for modeling glacier responses to ice shelf disintegration in order to improve our understanding of the process. Datasets generated for this study, such as surface velocity fields, DEMs and glacier front positions, are available via the PANGAEA database (https://www.pangaea.de), doi:10.1594/PANGAEA.859255.

Acknowledgments

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) in the framework of the priority programme "Antarctic Research with comparative investigations in Arctic ice areas" by a grant to M.B. (BR 2105/9-1). MB and TS would like to thank the HGF Alliance “Remote Sensing of Earth System Dynamics” for additional support. SM and JS were funded by IAA-DNA, project INST N°15. Exchange and field trips were partly funded by the European Commission 7th FP IRSES IMCONet project (FP7 IRSES, action no. 319718). JW and CR, acknowledge support from the Netherlands Polar Program of the Netherlands Organization for Scientific Research, section Earth and Life Sciences (NWO/ALW/NPP). IAA-DNA and AWI kindly provided logistic support as well as

68 Study 2: Dynamic response of Sjögren Inlet glaciers, Antarctic Peninsula, to ice shelf breakup derived from multi-mission remote sensing time series airborne operations. The authors particularly acknowledge the support of Marambio station personnel, the helicopters and Twin Otter DHC-6 crews for their support during the field surveys. Access to satellite data was kindly provided by various space agencies, e.g. under ESA AO 4032, DLR TerraSAR-X Background Mission Antarctic Peninsula & Ice Shelves, TSX AO LAN0013, TDX AO XTI_GLAC0264, ASF, GLIMS as well as NASA and USGS.

69 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

9 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Seehaus Thorsten1*, Cook Alison2, Barbosa Aline3, van Wessem Jan Melchior4, Reijmer C. H.4, Marinsek Sebastián5, Braun Matthias1

1 Institute of Geography, Friedrich-Alexander-University Erlangen-Nuremberg, Wetterkreuz 15, D- 91058 Erlangen, Germany 2 Department of Geography, Durham University, United Kingdom 3 Laboratório de Monitoramento da Criosfera, Fundação Universidade Federal do Rio Grande, Brazil 4 Institute for Marine and Atmospheric Research Utrecht, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands 5 Instituto Antártico Argentino, Balcarce 290, C1064AAF, Buenos Aires, Argentina To be submitted to The Cryosphere

Abstract

The climatic conditions along the northern Antarctic Peninsula showed significant changes within the last 50 years. Most studies focused on the glaciers affected by the disintegration of the ice shelves along the east coast. However, temporally and spatially detailed observations of the changes in ice dynamics along both coast lines are missing. Temporal trends of glacier area and ice surface velocity changes are derived from a broad multi-mission remote sensing database for 74 glacier basins on the northern Antarctic Peninsula (<65°S). By including ice thickness and climatic mass balance modeling data, ice discharge and glacier mass balances are computed. A recession of the glaciers by 238.81 km² is found for the period 1985-2015, whereat the glaciers affected by ice shelf disintegration showed the highest retreat. Along the west coast the average flow speeds of the glaciers increased by 41.5%. The former ice shelf tributaries on the east coast accelerated by 16.8% between 1992 and 2014 and similar temporal trends in ice velocities and area change are observed. However, the glaciers along the west coast revealed a strong spatial variability of the changes in ice dynamics, which are associated with geometric parameters of the individual glacier basin. The negative mass balance on the east side decreased from -5.02±3.86 Gt/a in the period 1992-2014 towards a nearly balanced state of -0.74± 1.7 Gt/a in the period 2010- 2014. The very high accumulation rates along the west coast compensate the increased glacier flow, leading to positive glacier mass balances. Consequently, for the period 2010-2014 balanced or positive mass budget of the whole study region are obtained, depending on the applied ice thickness and climatic mass balance scenario. Keywords: Antarctic Peninsula, remote sensing, glacier change, glacier dynamics, mass balance

70 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 9.1 Introduction During the last century, the Antarctic Peninsula (AP) has undergone significant warming (Carcass et al., 1998; Turner et al., 2005), leading to substantial glaciological changes. Numerous ice shelves along the AP (e.g. Larsen A/B, Prince Gustav and Wordie) have retreated widely or disintegrated in recent decades (Cook and Vaughan, 2010). As a consequence to the reduced buttressing, former tributary glaciers showed increased ice discharge and frontal retreat (e.g. De Angelis and Skvarca, 2003; Rack and Rott, 2004; Rignot et al., 2004; Seehaus et al., 2015; Wendt et al., 2010). Seehaus et al. (2015, 2016) quantified the ice loss of former ice shelf tributaries. Mass loss rates of -2.14±0.21 Gt/a (1995-2014) and -1.16±0.16 Gt/a (1993-2014) were found at Dinsmoor-Bombardier-Edgeworth Glacier System and Sjögren-Inlet glaciers, respectively. Glaciers that were not terminating in an ice shelf also showed considerable changes. Cook et al. (2005, 2014) have analyzed the variations of tidewater glacier fronts since the 1940s. The authors reported that 90% of the observed glaciers retreated, which they partly attributed to atmospheric warming. A more recent study revealed a mid-ocean warming along the southwestern coast of the AP, forcing the glacier retreat in this region (Cook et al., 2016). Pritchard and Vaughan (2007) observed an acceleration of ice flow by ~12% along the west coast of the AP (1995-2005) and linked it to frontal retreat and thinning of the tidewater glaciers. Observations by Kunz et al. (2012) support this assumption. The authors analyzed surface elevation changes at 12 glaciers on the western AP based on stereoscopic digital elevation models (DEM) over the period 1947-2010. Frontal surface lowering was found at all glaciers, whereas, area-wide surface lowering was observed on the north-eastern AP by various author groups (e.g. Berthier et al., 2010; Rott et al., 2014; Scambos et al., 2004; Wuite et al., 2015) as a consequence to ice shelf disintegration Consequently, the various observations suggest that the ice masses on the AP significantly lose ice and are contributing to sea level rise. Shepherd compiled a comprehensive glacier mass balance database of the polar ice sheets. The authors reported a mass loss on the AP of -36±10 Gt/a for the period 2005-2010, which corresponds to 35% of the total mass loss of Antarctica. On the northern AP, a mass loss rate of -24.9±7.8 Gt/a was reported by Scambos et al. (2014) for the period 2003-2008, indicating that major ice mass depletion happened at the northern part of the peninsula, especially along the eastern side at glaciers affected by ice shelf collapses. Hence, the authors found a nearly four times higher mass loss on the east coast compared to glaciers on the western AP. A projection of sea level rise contribution by the AP ice sheet amounts to 7-16 mm sea-level equivalent by 2100 and 10-25 mm by 2200 (Barrand et al., 2013a). However, along the western AP and on the higher elevation regions an increase in snow accumulation in the late 20th century was derived from ice cores and climate models (e.g. Dee et al., 2011; Goodwin, 2013; Potocki et al., 2011), whereas van Wessem et al. (2016) obtained insignificant trends in precipitation. Thus, the glacier's response to climate change on the AP is not homogeneous. Previous studies often just cover a specific period or region or focus on one aspect of glacier change. Therefore, we study the glacier changes on the northern AP (<65°S) between 1985 and 2015 by analyzing various remote sensing datasets and modeling results in order to obtain methodically consistent and temporally detailed time series of ice dynamic trends of 74 glacier basins. Variations in ice

71 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 surface velocities are correlated to changes of ice front positions. The observations are individually discussed for the sub regions, considering the different atmospheric, glaciological and oceanic conditions and changes. Finally, ice mass discharge is computed and compared with climatic mass balance estimates derived from atmospheric modeling in order to evaluate the glacier imbalances. 9.2 Study site

Figure 9-1: Panels A&B: Location of study site on the Antarctic Peninsula. Panel C: Separation of study site in 3 sectors and retreat states of Prince-Gustav and Larsen-A ice shelves. Red lines: profiles at glacier front for velocity measurements and ice mass flux computations. Map base, Landsat LIMA Mosaic © USGS, coastlines and catchment delineations from SCAR Antarctic Digital Database 6.0.

The AP is the northern most region of Antarctica. It covers only 3% of the entire continent, but receives 13% of the total mass input (van Lipzig et al., 2004, 2002). The AP acts as an orographic barrier for the circumpolar westerly air streams leading to very high precipitation values on the west coast and on the plateau region as well as frequent foehn type wind situations on the east coast (Marshall et al., 2006). The foehn events are characterized by strong winds and high air temperatures. Consequently, the climatic mass balance (bclim) shows a strong gradient across the mountain chain (van Wessem et al., 2016). Nearly all glaciers on the AP are marine terminating and most of the glacier catchments extend up to the high elevation plateau regions. Usually the AP plateau is separated from the outlet glaciers by escarpments and ice-falls. Glaciers on the west

72 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 coast drain into the Bellingshausen Sea and on the east coast into the Weddell Sea. Since the 1980s, the ice shelves along the east coast have substantially recessed and disintegrated (Prince Gustav and Larsen-A in 1995, Larsen Inlet 1987-89) (Cook and Vaughan, 2010; Skvarca et al., 1999). Scambos et al. (2003) attributed the retreat and collapse of ice shelves to higher summer air temperatures and surface melt. The climate on the northern AP is maritime and it is the only region of Antarctica that experiences widespread surface melt (Barrand et al., 2013b; Rau and Braun, 2002). Our study site stretches about 330 km from the northern tip of the AP southwards until Drygalski Glacier on the east coast and Grubb Glacier on the west coast (Figure 9-1). This narrow mountain chain covers an area of ~11.000 km² with altitudes stretching from sea level up to 2220 m. The glacier basins delineations are based on the Antarctic Digital Database ADD 6.0 (Cook et al., 2014). Glacier names are taken from the Global Land Ice Measurements from Space (GLIMS) project database. The local GLIMS glacier IDs (e.g. TPE62, LAB2) are used for unnamed glaciers. Still remaining missing glacier basin names are substituted by the ADD 6.0 glacier IDs. Some basins with lateral connected termini are merged. Names of the glaciers are merged as well (e.g. Sikorsky-Breguet-Gregory – SBG, see Table 9-1 for used abbreviations of glacier names). Due to the sparse data coverage, no time series analysis of the glaciers at the northern tip of the AP is possible. Therefore the northern most analyzed catchments are Broad-Valley Glacier on the east coast and TPE8 Glacier on the west coast, resulting in 74 studied glacier basins. Furthermore, the study region is divided in three sectors, taking into account the different climatic settings, drainage orientation and former ice shelf extend: sector “West” - Glaciers on the west coast, draining into the Bransfield and Gerlache Strait, sector “East” – Glaciers on the east coast, draining into the Prince Gustav Channel; sector “East-Ice-Shelf” – Glaciers on the east coast, that were former tributaries to the Larsen-A, Larsen-Inlet and Prince Gustav Ice Shelf. Table 9-1: Used abbreviations of glacier names Abbreviation Glacier names AMR Arago-Moser-Rudolph APPE Albone-Pyke-Polaris-Eliason CLM Cayley-Lilienthal-Mouillard DBE Dinsmoor-Bombardier-Edgeworth SBG Sikorsky-Breguet-Gregory

9.3 Data & Methods 9.3.1 Area Changes Changes in glaciers area are derived by differencing glacier outlines from various epochs. All observed glaciers are tidewater glaciers and only area changes along the calving front were considered. Information on the position of the glacier fronts at the study region are taken from Cook et al. (2014), and are available for the whole AP in the ADD 6.0 (1945-2010). This coastal-change inventory is based on manually digitizing of ice front positions using imaginary from various satellite sensors (e.g. Landsat, ERS) and aerial photo campaigns. This dataset is updated (up to 2015) and 73 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 gaps are filled by manually mapping of the ice front position based on SAR, Landsat and ASTER images. According to Cook et al. (2014) the ice-front positions are assigned to 5-year intervals in order to analyze temporal trends in glacier area changes in the period 1985-2015. Before 1985, only sparse information on ice front positions for the whole study region is available and the coverage by SAR data for analyzing glacier flow starts in 1992. The analysis of the area changes for the Larsen-A and Prince Gustav ice shelf tributaries is limited to the period 1995-2015, since the ice shelves disintegrated in 1995. The uncertainties of the glacier change measurements strongly depend on the used imaginary (e.g. spatial resolution, geodetic accuracies) as well as the methods used. To each record in the coastal-change inventory from the ADD 6.0 a reliability rating is assigned according to Ferrigno et al. (2006). The rating ranges from 1 to 5 (reliability within 60 m to 1 km) and takes into account errors due to manual digitization and interpretation (see Ferrigno et al., 2006 for a detailed description). This approach is also applied on the updated ice-front positions. Nearly all mapped ice fronts in the study region have a good reliability rating of 1 (76%) and 2 (21%). Only a few glacier fronts (3%) have a rating of 3. Reliability ratings of 4 and 5 are not applied. 9.3.2 Flow velocity Surface velocity maps are derived from repeat-pass Synthetic Aperture Radar (SAR) acquisitions. SAR image time series of the satellite missions ERS-1/2, Envisat, RadarSAT-1, ALOS, TerraSAR-X (TSX) and TanDEM-X (TDX) are analyzed, covering the period 1992-2014. Specifications of the SAR sensors are listed in Table 9-2. The huge amount of SAR images was provided by the German Aerospace Center (DLR), the European Space Agency (ESA) and the Alaska Satellite Facility (ASF). To obtain displacement fields of the glaciers, the widely used and well approved intensity offset tracking method is applied on co-registered single look complex SAR image pairs (Strozzi et al., 2002). In order to improve the co-registration of the image pairs, we mask out fast moving and unstable regions like outlet glaciers and the sea. Furthermore, single SAR image tiles acquired during the same satellite flyover are concatenated in the along-track direction. This helps to further improve the coregistration but also simplifies the analysis of the final results (no mosaicking of results needed). Image pairs with low quality coregistration are sorted out. A moving window technique is used by the intensity offset tracking method to compute the cross-correlation function of each image patch and to derive its azimuth and slant range displacement. Less reliable offset measurements are filtered out by means of the signal-to-noise ratio of the normalized cross- correlation function. Moreover, we apply an additional filter algorithm based on a comparison of the magnitude and alignment of the displacement vector relative to its surrounding offset measurements. This technique removes more than 90% of incorrect measurements (Burgess et al., 2012). Finally, the displacement fields are transferred from slant range into ground range geometry, taking into account the effects on the local incidence angle by the topography, geocoded, orthorectified and converted into velocity fields by means of the time span between the SAR acquisitions. The ASTER Global Digital Elevation Model of the Antarctic Peninsula (AP-DEM, Cook et al., 2012) is used as elevation reference. The mean date of the consecutive SAR acquisitions is assigned to each velocity field.

74 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 Table 9-2: Overview of SAR sensors and specifications used in this study. Intensity tracking parameters are provided in pixels [p] in slant range geometry. Platform Sensor Mode SAR Repetition Time interval Tracking Tracking band cycle [yyyy-mm-dd] patch sizes step size [d] [p x p] [p x p] ERS-1/2 SAR - C band 35/1 1992-12-08 48x240 5x25 2010-04-02 64x320 RADARSAT 1 SAR ST C band 24 2000-09-10 48x192 5x20 2006-09-03 64x256 64x256 Envisat ASAR - C band 35 2003-12-05 32x160 5x25 2009-08-16 64x320 128x640 ALOS PALSAR FBS L band 46 2006-05-18 64x192 10x30 2011-03-17 96x192 128x384 TerraSAR-X SAR SM X band 11 2008-10-14 128x128 25x25 TanDEM-X 2014-12-22 256x256 512x512

Depending on the displacement rate, the tracking window size needs to be adapted. For the fast flowing central glacier sections larger window sizes are needed, since large displacements cannot be tracked by using small correlation patches. Small tracking window sizes are suitable for the slow moving lateral sections of the outlet glaciers. Wide parts of large tracking patches cover the stable area next to the glacier, which biases the tracking results towards lower velocities. Consequently, we compute surface velocity fields of the same image pairs for different correlation patch sizes in order to get the best spatial coverage. Table 9-2 shows the different tracking window sizes for each sensor. The results of each image pair are stacked by starting with the results of smallest tracking window size and filling the gaps with the results of the next bigger tracking window size. The accuracy of the velocity measurements strongly depends on the coregistration quality and the intensity offset tracking algorithm settings. The mismatch of the coregistration is quantified by measuring the displacement on stable reference areas, like rock outcrops and nunataks. By means of the Bedmap2 (Fretwell et al., 2013, p. 2) and ADD 6.0 rock outcrop masks, reference areas are defined and the median displacements magnitude of each velocity field is measured at these areas. The uncertainty of the tracking process is estimated according to McNabb et al. (2012) depending on image resolution, time interval and oversampling factor. The accuracy of the tracking algorithm is estimated to be 0.2 pixels. Both independent error estimates are quadratically summed to compute the uncertainties of the individual velocity fields. Profiles are defined (red lines in Figure 9-1) close to the terminus of each glacier basin, taking into account the temporal changes of ice front position. The magnitude and direction (relative to north direction) of ice flow along the profiles of each tracking result are visually inspected in order to check the quality. Datasets with partial profile coverage or large data gaps as well as datasets with still remaining tracking errors are rejected. The changes in the ice flow of each glacier are analyzed by measuring the surface velocities within a buffer zone of 200m along the profiles and plotting the 75 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 temporal trend of the median values. The glaciers are classified in six categories according to the temporal evolution of the ice flow speeds (see Table 9-3). Only at glaciers with more than two observations and an observation period of more than 10 years the temporal trends are categorized. The GAMMA Remote Sensing software is used for processing of the SAR data. Table 9-3: Description of velocity change categories. *Ratings used for cluster analysis Category Description Rating* positive Long term increase of flow speed 2 peak Increase of flow speed with subsequent deceleration 1 stable Variability of measurements < 0.25 m/d 0 fluctuating Short term speed-ups and deceleration, no clear 0 trend trough Decrease of flow speed with subsequent -1 acceleration negative Long term decrease of flow speed -2

9.3.3 Ice mass flux Ice mass fluxes of the glaciers are calculated through the profiles at the glacier terminus defined in Section 9.3.2. An average ice density of 900 kg/m³ is applied. Ice thickness and its accuracy at the flux gates are extracted from the Huss and Farinotti (2014) ice thickness dataset of the AP. The flux gates are close to the grounding line positions (grounding line from MODIS Mosaic of Antarctica MOA, Scambos et al., 2007), hence basal friction is assumed to be very low and the surface velocities obtained by SAR intensity tracking are taken as mean flow velocities of the ice column. Surface velocity fields obtained from the modern high-resolution SAR sensors ALOS PALSAR and TerraSAR-X/TanDEM-X have got the best accuracies (Table 13-5, supplement) and spatial data coverage of the flux gates. Both SAR missions were launched in the second half of the study period. Hence, no time series analysis of ice mass flux is conducted and only the latest tracking results at each basin are used to compute the ice mass flux. The ice thickness reconstruction from Huss and Farinotti (2014) is based on surface velocity and climatic mass balance information as well as ice thickness measurements from various sources for model calibration and quality assessment. Most of the input data was acquired after 2005. Therefore, this dataset is used for the flux calculations based on the intensity tracking results at the end of the observation period (2010- 2014). The ice mass flux of each catchment at the beginning of the velocity time series (1992- 1996) is estimated by scaling the measured flux at the end of the study period by the observed median velocity values in the first and last year of the velocity time series (Section 9.3.2). Moreover, variations in ice thickness due to surface elevation changes are also considered in the flux estimates by integrating data from literature values. In sector “East” and “West”, surface elevation change values obtained by Scambos et al. (2014) in regions below 1000 m a.s.l. are applied. At DBE glaciers and Prince-Gustav Ice Shelf tributaries findings from Seehaus et al. (2015, 2016) are used and at the other glaciers in sector “East-Ice-Shelf” the observed lowering rates from Rott et al. (2014) are utilized.

76 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 9.3.4 Glacier attributes Glacier velocities and area change measurements provide information on the ice dynamics of the individual glaciers. To facilitate a better and comprehensive interpretation of these observations additional attributes considering the different geometries and settings of the glaciers are derived. Mass input strongly affects the dynamics of a glacier. The climatic mass balance at the northern AP shows a strong spatial variability, with very high accumulation rates along the west coast and significantly lower values on the east coast (see Section 9.4.6). But along each coast, bclim also shows an increase towards higher altitudes (see Section 9.5.2). Consequently, the mass input depends on the elevation range and the hypsometry. For each glacier basin a Hypsometric Index (HI), defined by Jiskoot et al. (2009), is calculated by means of surface elevations from the AP- DEM. Based on this index the glaciers are grouped into five categories ranging from very top- heavy to very bottom heavy (Table 9-4). Moreover, the maximum elevations of the individual glacier catchments are derived from the AP-DEM, which represents the altitude range of the catchment, since all observed glaciers are marine terminating. In order to quantify the flow pattern of the glacier catchments, the ratios (FA) of the flux gate cross sections divided by the glacier catchment areas are calculated. Lower values indicate that the outflow of the catchment is more channelized. Table 9-4: Hypsometric categories based on the Hypsometric Index according to Jiskoot et al., (2009) Hypsometric Index (HI) Hypsometric categories HI < -1.5 Very top-heavy -1.5 < HI < -1.2 Top-heavy -1.2 < HI < 1.2 Equidimensional 1.2 < HI < 1.5 Bottom-heavy HI > 1.5 Very bottom-heavy 9.3.5 Cluster analysis In order to study the influence of the glacier geometries on the glaciological changes and to find similarities, a cluster analysis for the glaciers in the sector “West” (Figure 9-1, red shaded area) is carried out. Variables of the glacier dynamics are the derived area changes (in percent) and velocity changes (ratings of the categories, Table 9-3). The glacier geometry parameters are the

Hypsometric Indexes HI, maximum surface elevation hmax of the basin and the flux gate to catchment size ratio FA. The variables are standardized before calculating the Euclidean dissimilarities between the observations. A hierarchical cluster analysis is applied on the dissimilarities using Ward's minimum variance method. 9.3.6 Climatic mass balance The regional atmospheric climate model RACMO2.3 at a horizontal resolution of 5.5 km (van Wessem et al., 2016) provides currently the most reliable climatic mass balance (defined by Cogley et al., 2011: surface mass balance + internal accumulation) dataset for the AP (Seehaus et al., 2016) covering the period 1979-2014. ERA-Interim reanalysis data (Dee et al., 2011) is used to force the model at the lateral boundaries. The climatic mass balance at the AP is dominated by precipitation and shows strong spatial variability, caused by the extreme elevation gradients and upwind-downwind situations (Turner, 2002). Even the high resolution RACMO2.3 model cannot 77 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 resolve all of these effects and topographic features. About 20% of the observed glaciers are smaller than a model grid cell (~30 km²) and about half of all basins are smaller than two model grid cells. Consequently, bclim is not derived for the individual catchments, but for the sectors of the study region. By averaging over these larger areas, the effects of the high spatial variability of the atmospheric parameters will be less significant.

9.4 Results 9.4.1 Area changes

Figure 9-2: Temporal trend of surface velocity (red) and area (blue) changes of selected glaciers in the study region.

Area changes relative to the measurements in the epoch 1985-1989 of selected glaciers are plotted in Figure 9-2 and of all observed glaciers in Figure S1-S74 (supplement). The glaciers are classified in three groups based on the latest area change measurement: retreated (Figure 9-2A, B, C, F) – loss of glacier area by frontal retreat; stable (Figure 9-2E) – no significant area changes (within the error bars); advanced (Figure 9-2D) – gain of glacier area by frontal advance. In Figure 9-3 the spatial distribution of the area change classification is illustrated. All glaciers along the east coast, including the former ice shelf tributaries, retreated, whereas along the west coast, numerous glaciers show stable ice front positions and some glaciers even advanced. In total -238 km² of glacier area was lost at the study site in the period 1985-2015, which corresponds to a relative loss of 2.2%. All sectors show glacier area loss (Table 9-5), whereof the area loss at sector “East-Ice- Shelves” clearly dominates. The temporal trends of total glacier area and area loss of all observed glaciers and of each sector are presented in Figure 9-4.

78 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Figure 9-3: Categorizations of glaciers according to temporal trends in area changes (dots) and flow velocities (symbols). Colors of catchment delineation indicate Hypsometric categories according to Jiskoot et al. (2009).Background: Landsat LIMA Mosaic © USGS

79 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Figure 9-4: Glaciers area (gray bars) of the whole study site (Panel A) and of the individual sectors (Panels B-D) in the period 1985-2015. Changes in glacier area (blue points) are relative to the measurements in time interval 1985-1990. Note the different scaling of the y-axes. In sector "East", area changes before 1995 are only measured at Larsen-Inlet tributaries (APPE glaciers).

9.4.2 Flow velocity In total 282 stacked and filtered velocity fields are derived from the SAR acquisitions covering the period from 25th December, 1992 until 16th December, 2014. The average total uncertainty of the velocity fields amounts to 0.08 m/d. In Table 13-5 (supplement) the error estimates of each velocity field are listed. The mean sample count to estimate the coregistration quality is 11717 and the average mismatch amounts to 0.07 m/d. The error caused by the tracking algorithm strongly varies depending on the source of the SAR data (sensor) and a mean value of 0.05 m/d is found. ERS image pairs with time intervals of one day have very large tracking uncertainties, due to the very short temporal baselines, and are not considered in the total error computations. All measured velocity profiles of the 74 observed glaciers are visually inspected and finally 2503 datasets passed the quality check (on average ~34 per glacier). Figure 9-2 shows exemplarily the temporal evolution of the ice flow for each velocity change category (see Table 9-3). The temporal trends of the surface velocities at the termini of each glacier are plotted in Figures S1-S74 (supplement) and the related category is listed in Table 13-6 (supplement). The spatial distribution of the categories is illustrated in Figure 9-3. At nearly all glaciers in sector “East-Ice-Shelf” a peak in ice velocities is observed. In the sector “East”, most glaciers showed a decrease in flow velocities in the observation period. The glaciers on the west coast show a more irregular distribution than along the east coast, but some local clustering can be observed. 80 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Figure 9-5: Spatial distribution of glacier types in sector “West”. Glaciers are group based on a hierarchical cluster analysis. Catchment colors: Measured ice mass ice flux in the period 2010- 2014.

81 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 The mean velocities in the first and in the last year of the observation period of all analyzed glaciers and for each sector are also listed in Table 9-5. For each glacier the mean values as well as the absolute and relative change is presented in Table 13-6 (supplement). On average the ice flow in the study region increased by 1.6%. Along the west coast an average acceleration by 41.5% and, at the former ice shelf tributaries, by 16.8% is found. In the sector “East” the glaciers decelerated resulting in a mean velocity change of -69%. The shortest observation period is 14.8 years at DBC31 Glacier and on average velocity changes are analyzed over a period of 19.4 years.

Table 9-5: Summary of observed parameters for each sector and all glaciers. lf – length of ice front; A – Glacier area in the respective period; dA – Change in glacier area between 1985 and 2015; dt: mean time period of velocity measurements, vs – mean of earliest velocity measurements (1992-1996); vE – mean of latest velocity measurements (2010-2014), dv – mean velocity change; nv – sum of velocity measurements in the observation period (dt); FS – Ice mass flux at the time of earliest velocity measurements (1992-1996); FE – Ice mass flux at the time of latest velocity measurements (2010-2014); ti – mean ice thickness at the flux gates, hmax – average maximum altitude of individual basins, bclimS – means climatic mass balance in the period of earliest flux estimates (1992-1996); bclimE – mean climatic mass balance in the period of latest flux estimates (2010-2014); bclim – mean climatic mass balance in the period 1979-2014; IS - Imbalance in the period 1992-1996 (bclimS - FS); IE: Imbalance in the period 2010-2014 (bclimE - FE) Sector East East-Ice-Shelf West All glaciers

lf [m] 85114 127909 268763 481786 A1985-1990 [km²] 1538.78 3655.13 5809.33 11003.23 A2010-2015 [km²] 1517.71 3446.54 5800.18 10764.42 dA [km²] -21.07 -208.59 -9.14 -238.81 dt [y] 18.79 19.05 20 19

vS [m/d] 0.995 0.480 0.427 0.537

vE [m/d] 0.307 0.561 0.605 0.545 dv [m/d] -0.688 0.081 0.177 0.008

nv 319 584 1600 2503

FS [Gt/a] 5.59±2.91 4.55±0.95 11.20±4.56 21.94±8.54

FE [Gt/a] 1.99±0.87 4.29±0.83 11.65±4.52 17.93±6.22 ti [m] 240±92 263±70 189±65 211±71

hmax [m] 1339 1891 1636 1629

bclimS [kg/(m²a)] 1093 920 4238 2831

bclimE [kg/(m²a)] 1246 1023 3949 2717

bclim [kg/(m²a)] 1119 939 3769 2573

IS [Gt/a] -3.92±2.91 -1.28±0.95 14.02±4.56 8.06±8.54

IE [Gt/a] -0.09±0.87 -0.65±0.83 11.85±4.52 10.86±6.22 9.4.3 Ice mass flux Figure 9-5 shows the spatial distribution of the ice flux. In sector “East” a strong decline in mass discharge by 64% and in sector “East-Ice-Shelf” a slight recession by 6% is observed (Table 9-5). In the sectors “West”, an increase in ice mass flux by 4% is obtained. For the whole study region a total ice discharge of 21.94±8.54 Gt/a in the period 1992-1996 and of 17.93±6.22 Gt/a in the period 2010 – 2014 is found, corresponding to a reduction of the ice flux by -18%. Ice discharge of the catchments at the beginning and end of the velocity time series years and the respective dates are listed in Table 13-6 (supplement).

82 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 9.4.4 Glacier attributes The spatial distribution of Hypsometric Indexes and categories of the glacier basins is presented in Figure 9-3 and the values are listed in Table 13-6 (supplement). No clear distribution pattern can be identified, reflecting the heterogeneous topography of the AP. The maximum elevation of the catchments and the FA factors are also listed in Table 13-6 (supplement). 9.4.5 Cluster analysis

Figure 9-6: Dendrogram of hierarchical cluster analysis of glaciers in sector "West". The glaciers are assorted in four groups (red rectangles)

Figure 9-7: Boxplots of cluster analysis input variables (Sector “West”) for each group. Whiskers extend to the most extreme data points. 83 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 The resulting dendrogram of the hierarchical cluster analysis is plotted in Figure 9-6. Four groups are distinguished. The boxplots of each input variable are generated based on this grouping and are shown in Figure 9-7. The characteristics of the groups are discussed in Section 9.5.1.3. 9.4.6 Climatic mass balance The annual climatic mass balance (per calendar year) of the study region and of each sector is plotted in Figure 9-8 in the period 1997-2014. An insignificant long term increase of bclim by 0.09±0.08 Gt/a and a mean climatic mass balance of 2573 kg/(m²a) is found for the observed area.

A nearly four times higher bclim is observed in the sector “West” as compared to the other sectors (Table 9-5). Average climatic mass balances are also calculated for the periods 1992-1996 and 2010-2014 (Table 9-5), which facilitates the comparison with the obtained ice mass fluxes.

Figure 9-8: Temporal changes of climatic mass balances (bclim; Bclim = bclim*Area) of the whole study site (Panel A) and of the individual sectors (Panel B-D) from the RACMO2.3 model (5.5 km horizontal resolution) in the period 1979-2014. Linear curve fittings (red lines) are applied to estimate the trend in Bclim.

9.5 Discussion 9.5.1 Ice dynamics Most of the observed glaciers (62%) retreated and only 8% advanced in the study period. These findings are comparable to the results of Cook et al. (2005, 2014, 2016). Only glaciers along the west coast showed stable or advancing calving fronts and all glaciers on the east coast recessed.

84 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 This heterogeneous pattern was also found by Davies et al. (2012) at Western Trinity Peninsula. Most significant retreat occurred in the sector “East-Ice-Shelf”. In the period 1985-1995, the Larsen-Inlet tributaries (APPE glaciers) lost 45.0 km² of ice. After the disintegration of Prince- Gustav and Larsen-A Ice shelf, the tributaries rapidly retreated in the period 1995-2005. The recession slowed down in the latest observation interval (2005-2010). This trend is comparable to detailed observations by Seehaus et al. (2015, 2016) at individual glaciers (DBE glaciers and Sjögren-Inlet glaciers). At sector “East” the highest area-loss is found in the earliest observation interval (1985-1990). Davies et al. (2012) also reported higher shrinkage rates for most of the glaciers in this sector in the period 1988-2001 than in the period 2001-2009. Moreover, increased recession is also found in the time period (1995-2005, Figure 9-4) at sector “East”. It could be indirectly caused by the disintegration of Larsen-A and Prince-Gustav ice shelves. Davies et al. (2012) and Hulbe et al. (2004) supposed that the disintegration of an ice shelf affects the local climate. The air temperatures would rise due to the presence of more ice free water in summers. At Base Marambio, ~100 km east of this sector, about 2°C higher mean annual air temperatures were recorded in the period 1997-2007 (Seehaus et al., 2015). Unfortunately, no temperature records are available in sector “East” covering this period. The average changes of flow velocities at each sector also vary strongly (Table 9-5) in the observation period 1992-2014. On the west coast an increase of 42% is found, whereas in sector “East” the glaciers slowed down by about 69% and at the ice shelf tributaries the ice flow increased on average by 17%. Pritchard and Vaughan (2007) reported an increase in mean flow rate of 7.8% in frame 4923 (the central and northern part of sector “West”) and 15.2% in frame 4941 (the southern part of sector “West”) for the period 1992-2005. This spatial trend corresponds to our observations, since most of the glaciers with a clear positive velocity trend are located at the southern end of sector “West”. Nevertheless, our observed accelerated ice flow on the western side is 3.5 times higher than their findings. Our study period reaches until 2014. Assuming a linear acceleration, the findings of Pritchard and Vaughan (2007) would result in an increased ice flow of ~20%. Furthermore, they estimated the mean velocity change by measuring the flow speed at profiles along the flow direction of the glacier, whereas we measured at across glacier profiles at the terminus. If a tidewater glacier speeds-up due to the destabilization of its front, the far highest acceleration is found at the terminus (see Seehaus et al., 2015, Figure 7-3). Additionally, Pritchard and Vaughan (2007) just analyzed the azimuth components of the SAR tracking results (in order to improve the accuracy), but the main flow direction of most of the glaciers in sector “West” is preferable in range direction of the used SAR acquisitions. Therefore, the azimuth components of the flow vector are less affected by the glacier speed up. Consequently, the difference in the applied measurement methods explains the discrepancies between both studies. In the following section the observed changes in the individual sectors are discussed in more detail. 9.5.1.1 East-Ice-Shelf In the sector “East-Ice-Shelf” nearly all glaciers showed a rapid and significant acceleration after ice shelf break up and a subsequent slow down (Figures S14-S26, supplement). A quite gentle peak in flow speeds is obtained at LAB32 and TPE114 glaciers. They are classified as “stable”,

85 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 since the variations are below the threshold of 0.25 m/d, according to the categorization in Table 9- 3. Dramatic speed up with subsequent deceleration of former ice shelf tributaries was reported by various authors, e.g., Seehaus et al., (2015, 2016) in this sector (at DBE and Sjögren-Inlet glaciers, but measured at different locations) and further south at Larsen-B embayment by Rott et al. (2011) and Wuite et al. (2015). The velocities reported by Rott et al. (2014) at Sjögren, Pyke, Edgeworth and Drygalski glaciers are generally higher than our findings. The authors measured at locations near the center of the glacier fronts, where the ice flow velocities are typically highest, whereas we measured the median velocities at across profiles close to the glacier fronts. The different approaches result in different absolute values but equal temporal trends are observed by all measurement techniques. Highest peak values of 6.3 m/d are found at TPE61 Glacier in November 1995 and January 1996. Most glaciers (Arron Icefall, Drygalski, LAB2, TPE61, TPE62) decelerated towards pre-collapse values and show nearly constant flow speeds in recent years, indicating that the glaciers adjusted to the new boundary conditions. At “2558”, Boydell, DBE and Sjögren glaciers the deceleration is ongoing and Boydell and DBE glaciers still show increased median flow speeds at the glacier fronts. Thus, these tributary glaciers are still adjusting to the new boundary conditions, as suggested by Seehaus et al. (2015, 2016). In the 1980s, Prince Gustav Ice Shelf retreated gradually (see Figure 9-1) and “2668” Glacier has not been buttressed by the ice shelf since the early 1990s. A deceleration is found in the period 2005-2010 (Fig. S15, supplement). Hence, this glacier might also have experienced a speed up in the early 1990s due to the recession of Prince Gustav Ice Shelf in the 1980s. However, the earliest velocity measurement at “2668” Glacier is available from February 1996. The Ice shelf in Larsen-Inlet disintegrated in 1987-1988 and earliest velocity measurements are obtained for 1993. Therefore, a potential peak in the flow speed after ice shelf brake-up cannot be detected at APPE glaciers (Fig. S16, supplement). The ice flow has shown constantly short term variations in the order of 0.2-0.5 m/d but no clear long term trend since 1993. Rott et al., (2014) also found nearly constant flow velocities at Pyke Glacier. The authors suggest, that the ice flow of APPE glaciers were not strongly disturbed by the ice shelf removal, due to the steep glacier surfaces and quite shallow seabed topography at the glacier fronts (Pudsey et al., 2001). 9.5.1.2 East The glaciers north of the former Prince-Gustav Ice shelf show in general a trend towards lower flow velocities. Eyrie, Russell East, TPE31, TPE32, TPE34, TPE130 and “2731” glaciers experienced a rapid decrease and all, except “2721” Glacier, a subsequent stabilization or even gentle acceleration of flow velocities. A significant retreat followed by a stabilization or slight re-advance of the calving front position is also observed at these glaciers. According to Benn and Evans (1998), a small retreat of a glacier with an overdeepening behind its grounding line (bed sloping away from the ice front) can result in a rapid recession into the deepening fjord. The increased calving and retreat of the ice front cause stronger up-glacier driving stress, higher flow speed as well as glacier thinning and steepening (Meier and Post, 1987; Veen, 2002). The glacier front stabilizes when the grounding line reaches shallower bathymetry and ice flow also starts to slowdown. A delay between the front stabilization and slowdown can be caused by thinning and steepening of the glacier. Additionally, the accelerated ice flow can excel the retreat rates and cause short-term

86 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 glacier advances in the period of high flow speeds (e.g. Eyrie, Russel East, TPE32 and TPE130 Glacier) (Meier and Post, 1987). The discussed process can be initiated by climatic forcing (Benn and Evans, 1998). Skvarca et al. (1998) reported a significant increase in surface air temperatures at the north-eastern AP in the period 1960-1997 and correlated it with the recession of the Larsen and Prince-Gustav Ice shelves and the observed retreat of tidewater glaciers on James Ross Island in the period 1975-1995 (Skvarca et al., 1995). Hence, we assume that the initial recessions of the glaciers in sector “East” were forced by changing climatic conditions and happened before 1992 (earliest velocity measurements available). Therefore, we observe quite high flow speeds at the beginning of our ice speed time series. The fjord geometry affects the dynamics of the terminus of a tidewater glacier (Benn and Evans, 1998; Veen, 2002). The tongues of Aitkenhead and “2702” glaciers are split in two branches by nunataks, resulting in rather complex fjord geometries. A retreat from pinning points (e.g. fjord narrowing) causes further rapid recession and higher flow speeds until the ice front reaches a new stable position as observed at Aitkenhead Glacier. The “2702” Glacier also shows faster ice flow during its retreat followed by a front stabilization and deceleration. At TPE10 Glacier a “peaked” flow velocity trend is observed as at Aitkenhead Glacier. No rock outcrop is visible at the terminus, but a bedrock bump (north of the center of the ice front, identified by visual inspection of optical satellite imagery) acts as a pinning point. Most probably, it supports the glacier tongue after the discovered retreat. The front of Broad Valley Glacier is located in a widening fjord. This geometry makes the glacier less vulnerable to frontal changes (Benn and Evans, 1998). Therefore, no significant changes in flow velocities are observed as a consequence of the frontal recession and re-advance. Diplock and Victory glaciers show a decrease of flow speed during retreat and an acceleration combined with frontal advance. This behavior is similar to surge-type glaciers found for example in the Karakorum or on Svalbard (e.g. Rankl et al., 2014; Strozzi et al., 2000). Surge-type glaciers are characterized by episodic rapid down-wasting, resulting in a frontal acceleration and strong advance. At tidewater glaciers the advance can be strongly compensated by increased calving rates in deepwater in front of the glacier. Thus, both glaciers could have experienced a surge cycle in our observation period; however, a longer time series analysis is necessary to proof this hypothesis. 9.5.1.3 West Meredith and King (2005) reported an increase of surface summer temperatures by more than 1°C in the ocean west of the AP. The authors attributed it to atmospheric warming and reduced sea ice production rates. However, Cook et al. (2016) reported a cooling in the sea along the north-western AP for the period 1945-2009, and an absence of the warming at the AP since the turn of the millennium was found by Turner et al. (2016). Consequently, the atmospheric warming in the 20th century, led to the observed glaciological changes in the sector “West”. Along the west coast the changes in glacier area and flow speed show a much more heterogeneous pattern as compared to the eastern sectors. Especially in the southern part of the study region, the coastline is dominated by a fractal structure. Kunz et al. (2012) obtained thinning at the glacier termini along the western AP. An acceleration and terminus retreat can be caused by

87 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 these processes as shown by Benn et al. (2007). However, the authors also point out that changes in ice thickness not necessarily affect the ice flow and that calving front position and ice dynamics strongly dependent on the fjord and glacier geometries. The large number of glaciers in this sector is analyzed by means of a hierarchical cluster analysis (Section 9.3.5) and assorted in four groups based on the dendrogram (Figure 9-6). The termini of glaciers in group 1 are typically located in narrow fjords (Figure 9-5) and are clustered in Charcot, Charlotte and Andvord Bay, whereas glaciers of group 2 are more spread all over the study region, with a local clustering in Wilhelmina Bay. Glaciers of group 3 and 4 usually face the open ocean, and do not terminate in fjords (especially in the northern part, Trinity Peninsula). Boxplots of the individual variables of each group are shown in Figure 9-7 and the observed glacier dynamics are discussed in the following. Group1 (14 glaciers): Most glaciers experienced acceleration. The majority of the glacier basins are “very top-heavy” or

“top-heavy” (median HI = -1.8), stretching from sea level up to 1892 m in average. The bclim increases toward higher altitudes (Figure 9-9) and highest values are found in regions between 1000 and 1700 m a.s.l. Consequently these glaciers receive high mass input in their large high altitude regions. The accumulation on the AP has increased by 20% since the 1950s (Peel, 1992) and also a slight increase in bclim is observed at the study region in the period 1979-2015 (Figure 9- 8). Pritchard and Vaughan (2007) reported that only a small fraction of the acceleration can be attributed to a glacier thickening due to increased mass input. Up-glacier thickening (observed by Kunz et al., 2015), due to an increase in accumulation in higher altitudes, in combination with the observed frontal thinning, leads to a steepening of the glacier and an increase in driving stress, resulting in faster ice flow (Meier and Post, 1987). Moreover, a thinning of the terminus reduces the effective basal stress of a tidewater glacier and facilitates faster ice flow (Pritchard and Vaughan, 2007). The flux gate cross sections to catchment size ratios are relatively small, indicating narrowing catchments towards the ice front. The channelized increased ice flow nearly compensates the increased calving rates (due to frontal thinning), resulting in an average shrinkage of the glaciers by only 0.2% in the period 1985-2015. The high flow speeds can even compensate the calving and contribute to ice-front advances (Krebs and TPE46 Glacier). Group 2 (19 glaciers)

Group 2 shows similar hmax and FA characteristics as group 1. Area changes are also quite small (- 0.1%). Most of the glaciers experienced positive or “peaked” velocities trends. In contrast to group 1 the catchments are all “bottom-heavy” and some are even “very bottom-heavy”. We assume that the constraints are similar to group 1 (increasing bclim, frontal thinning, steepening). However, the glaciers are affected differently because of their “bottom-heavy” geometries. The additional mass accumulation in the upper regions is smaller. Consequently, the imbalance due to the frontal thinning and up-glacier mass gain is less pronounced as in group 1 and numerous glaciers (“peak” type) started to decelerate after the speed-up, indicating that these glaciers are adjusting to the new boundary conditions.

88 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Figure 9-9: Hypsometry of climatic mass balance (bclim) for the whole study site and the individual sectors extracted from the RACMO 2.3 model (5.5 km horizontal resolution).

Group 3 (13 glaciers) These basins show typically a “bottom-heavy” hypsometry and smaller elevation ranges (in average up to 1103 m a.s.l.). Thus, bclim is relatively low (see Figure 9-9). Frontal thinning does not significantly affect the ice flow, probably due to flat glacier topography and the low mass input. Consequently, the flow speed is in general stable or even slightly decreases in the observation period. The smaller mean ice thickness at the termini (161 m, compared to 211 m of all glaciers) of group 3 implies less interaction with the ocean, leading to a quite small average area loss of ~0.1%. Group 4 (3 glaciers) All basins in this group have got a “very bottom-heavy” hypsometry and a comparable elevation rage as group 3 glaciers. Thus, this group shows similar trends in velocity changes as group 3. Additionally, the FA factors are in general even higher as in group 3, implying that outflow of the catchments is less channelized and the glacier fronts are relatively long compared to the catchment sizes. This explains why the largest relative area changes, in average -5.1%, are found at glaciers in group 4. 9.5.2 Mass balance The observed reduction in ice flux for the whole study region is strongly biased by the slowdown of the glaciers in the sector “East” (Table 9-5). As discussed above, most glaciers in this region probably experienced an epoch of glacier speed up before 1992, followed by the observed deceleration. Therefore, a strong flux regression is obtained in the study period (1992-2014). At sectors “West” and “East-Ice-Shelf” the changes in ice discharge are smaller than the error bars. Consequently, we conclude that the observed accelerated ice-flow is compensated by the thinning of the glacier tongues, keeping the ice mass flux nearly stable.

89 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 Seehaus et al. (2015, 2016) reported ice mass flux at DBE glaciers of 0.77±0.12 Gt/a in the period 1993-1996 and 0.73±0.07 Gt/a in 2014. At Sjögren-Inlet glaciers, they estimated the ice discharge to be ~0.4 Gt/a in 1996 and ~0.65 Gt/a in 2014 (sum of Boydell and Sjögren glaciers). The estimates in from both studies agree with our estimates within the error range (Table 13-6, supplement). These glaciers drain nearly one third of the sector “East-Ice-Shelf” and also their ice flux is about one third of the sector's total ice discharge. Drygalski Glacier is the largest catchment in this sector and also in the whole study site. It also shows the highest measured ice flux of 1.34±0.08 Gt/a in 2010, which is comparable to recent discharge rates of the high dynamic Crane, Hektoria, Flask and Green glaciers in the Larsen-B embayment (Rott et al., 2011; Wuite et al., 2015). The imbalance of the study region and the individual sectors is estimated by subtracting the ice discharge from Bclim (total climatic mass balance of the respective area). This approach is called the Input-Output method or also known as the Flux Gate method. The east side of the AP shows negative imbalances of -1.3 Gt/a and -3.9 Gt/a in the period 1992-1996 at sector “East-Ice-Shelf” and “East”, respectively (Table 9-5). In recent years, the imbalances are less negative. High accumulation rates on the west side exceed the mass discharge, albeit that a decrease in bclim is found in recent years. The positive imbalances at the west side surmount the mass loss along the east coast. Hence, positive imbalances of the study region are found for both epochs. Recent studies by Seehaus et al. (2015, 2016) revealed that the ice thickness reconstructions from Huss and Farinotti (2014) underestimate the ice thickness at the termini. It is currently the best available dataset at the AP. However, often a rise in the bed topography is observed towards the ice margin, resulting in too shallow ice thickness estimates at the ice fronts (Seehaus et al., 2015, 2016). The largest uncertainties in ice flux are caused by the errors of the ice thickness (see also Seehaus et al., 2015, 2016). Therefore, we calculated the imbalances by using the upper bound estimation (IU) of the ice mass flux (upper limit of error bar), in order to take into account the underestimation of the ice thickness. This approach leads to a slight negative imbalance of -0.5 Gt/a in 1992-1996 and a mass gain rate of 4.6 Gt/a in 2010-2014. Scambos et al. (2014) derived geodetic mass balances of the northern AP in the period 2001-2010 (primarily 2003-2008) by using satellite laser altimetry and stereo-images. They observed a total mass change of -24.9±7.8 Gt/a at regions north of 66°S (incl. surrounding islands) and -6.2±2.7 Gt/a in our study region. In 2002, the Larsen-B ice shelf disintegrated and subsequently the tributary glaciers reacted with highly increased ice discharge (e.g. Berthier et al., 2012; Rack and Rott, 2004, 2003; Rott et al., 2011; Shuman et al., 2011). In recent years, the glaciers significantly slowed down and the mass loss also decreased (Wuite et al., 2015). Therefore, the imbalance estimates of the whole northern AP presented by Scambos et al. (2014) are strongly dominated by this event. Nearly half of the reported mass loss can be attributed to the glaciers in the Larsen-B embayment (incl. Scar-Inlet tributaries). Concerning our study region, the former Larsen-A and Prince-Gustav Ice Shelf tributaries showed higher flow speeds (“peaked” velocity trends) and consequently higher mass discharge in the period 2001-2010 compared to the periods 1992-1996 and 2010-2014. This leads to the large differences between our imbalances estimates and findings by Scambos et al. (2014). Observations by Seehaus et al. (2015, 2016) support our conclusion.

90 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Our imbalance estimates for the west side are always positive, caused by the high bclim values.

Applying the average bclim=1921 kg/m²a found by Scambos et al. (2014) in sector “West” (Data from the Regional Atmospheric Climate Model RACMO-2.1/ANT, Lenaerts et al., 2012) leads to nearly balanced mass budgets of -0.06±4.52 Gt/a and -0.51±4.56 Gt/a in the periods 1992-1996 and 2010-2014, respectively, which overlap within the error bars with findings by Scambos et al. (2014) (-2.4±1.7 Gt/a). We summarize that the differences between our study and the work by Scambos et al. (2014) are on the one hand caused by the different observation periods. On the other hand, the different applied mass balance methods have certain limitations, e.g. deficiencies in ice thickness data for the Input-Output method or spatial interpolation of elevation change information for the geodetic method. Moreover, depending on the applied bclim data, the mass change estimates using the Input-Output Method can vary significantly. McMillan et al. (2014) computed mass change rates for whole Antarctica by using CryoSat-2 data in the period 2010-2013. Imbalances of -9±9 Gt/a and -4±9 Gt/a were obtained at the north-eastern and north-western basins of the AP, respectively. Sasgen et al. (2013) derived for the northern section of the AP an ice mass change rate of -26±3 Gt/a from GRACE data in the time interval January 2003 to September 2012. A mass change rate of -16.3±5.9 Gt/a is detected by Groh and Horwath (2016) at the northern AP in the period February 2010 till November 2014 (data available via https://data1.geo.tu-dresden.de/ais_gmb/). The observed regions in the studies named above are all much larger than ours. The authors used the IceSAT drainage basins (Zwally et al., 2012) for their estimates. These basins extend much further south. They include Marguerite Bay on the west side (basin 25) and the Larsen-B and Larsen-C embayments on the east coast (basin 26), consequently resulting in higher mass loss rates. The glaciers in Marguerite Bay were affected by the disintegration of Wordie Ice Shelf in the 1970s-80s and Rignot et al. (2005) estimated the imbalance to be ~3.1 km³/a (2.8 Gt/a) at Airy, Rotz, Seller and Fleming glaciers. A more recent work by Wendt et al. (2010) showed that the glaciers continued to thin (-4.1±0.2 m/a) and the flow speeds were still 40-50% higher in 2008 than in 1974. Ongoing studies also indicated that the former Wordie Ice Shelf tributaries still lose ice (Friedl P., personal communication). At Larsen-B embayment on the east side, the most recent imbalance estimates by Wuite et al. (2015) indicate a mass change rate of -3.2 Gt/a (sum of Crane, Jorum, Hektoria-Green, Flask and Leppard glaciers). Khazendar et al. (2011) detected a speed up Larsen-C Ice-Shelf and its tributaries, indicating an increased ice discharge. Moreover, the study region of Sasgen et al. (2013) also includes adjacent islands that lost ice, especially the South Shetland Islands (Osmanoglu et al., 2013, 2014). Gardner et al. (2013) estimated a mass change of -7.3±4.0 Gt/a on the AP Islands in the period 2003-2009. The glaciological and climatic changes along the AP lead to a high spatial and temporal variability of the mass balance. Therefore, we conclude that the comparison of results from different periods and covering different extents is difficult and needs to be carefully discussed. 9.6 Conclusions The glaciers along the east and west coast of the study region show different temporal trends in ice dynamics. Along the east coast all glacier fronts retreated in the period 1985-2015, with highest retreat rates observed at former ice shelf tributaries. Moreover, all glaciers affected by ice shelf

91 Study 3: Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985 disintegration show similar temporal trends of ice dynamics, whereas on the west side, no clear change pattern is obvious. The changes in ice flow and glacier area along the west coast could be associated with the geometries of the individual catchments, by applying a cluster analysis. The mass loss on the east coast decreased and a nearly balanced mass budget of this region is found for the period 2010-2014. Thus, the dynamic responses of the former tributary glaciers to ice shelf brake-up are calming down and the glaciers are adjusting to the new conditions, as suggested by Seehaus et al. (2015, 2016). The average flow speed of the glaciers along the west side significantly accelerated in the observation period (1992-2014). However, Kunz et al. (2012) hypothesized, that the high climatic mass balance along the west coast compensates the increase in ice discharge and reported surface lowering at the glacier tongues. Our observed positive mass balances along the west side support this hypothesis. Consequently, we estimate a balanced or positive mass budget of the whole study region for the period 2010-2014, depending on the ice thickness and climatic mass balance settings. Previous studies revealed negative mass balances on the northern AP, but they covered different observation periods and applied different methods, e.g. Scambos et al. (2014) observed geodetic mass balances in the period 2003-2008. Thus, the considerable decrease in mass loss of the former ice shelf tributaries and high accumulation rates along the west coast as well as the reported decrease in air temperatures (especially during austral summer) at the AP since the beginning of the 21st century (Turner et al., 2016) support our findings of balanced or positive mass budgets along the northern AP.

Acknowledgments

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) in the framework of the priority programme "Antarctic Research with comparative investigations in Arctic ice areas" by a grant to M.B. (BR 2105/9-1). MB and TS would like to thank the HGF Alliance “Remote Sensing of Earth System Dynamics” for additional support. JW and CR, acknowledge support from the Netherlands Polar Program of the Netherlands Organization for Scientific Research, section Earth and Life Sciences (NWO/ALW/NPP). IAA-DNA and AWI kindly provided logistic support as well as airborne operations. Access to satellite data was kindly provided by various space agencies, e.g. under ESA AO 4032, DLR TerraSAR-X Background Mission Antarctic Peninsula & Ice Shelves, TSX AO LAN0013, TDX AO XTI_GLAC0264, ASF, GLIMS as well as NASA and USGS.

92 Conclusions and Outlook

10 Conclusions and Outlook

Detailed studies on the changes in ice dynamics along the northern Antarctic Peninsula were carried out by exploiting the remote sensing data archives and planned data takes, acquired by a large variety of air and spaceborne sensors. Temporal trends of ice flow, glacier front position, surface elevation and mass balance could be analyzed in detail due to the high density of observations available. At the former ice shelf tributaries, significant speed-up and rapid glacier front recession after ice shelf disintegration and a subsequent deceleration and glacier front stabilization were observed. The variations in magnitude and timing of the adjustments were assigned to the individual geometric settings of the glaciers. Especially in the last decade, the increase of acquisitions on the Antarctic Peninsula and the improved sensor characteristics facilitated the observation of short- term changes and the detection of seasonal variabilities. Temporary speed-ups and retreat events in summers could be identified, which are probably caused by the weakening of the ice mélange in the fjords during summers, leading to a reduced buttressing of the glacier tongue. The mass balance computations revealed high mass losses subsequent to ice shelf break up, which decreased in recent years. The deceleration and front stabilization indicate that the glaciers are slowly adjusting to the new boundary conditions. However, the former tributary glaciers are still showing considerable surface lowering and negative imbalances even about 20 years after the collapse of the ice shelf. The projected recession of the Filchner-Ronne Ice Shelf or a collapse of the ice shelves in the Amundson Sea could lead to a destabilization of some of the biggest ice streams in Antarctica and dramatic contributions to global SLR (e.g. Dutrieux et al., 2014; Hellmer et al., 2012; Hughes, 1981). Furthermore, widespread retreat and thinning of ice shelves around the Antarctic continent and major rifts in the Larsen-C and SCAR Inlet ice shelves were observed by various authors (e.g. Jansen et al., 2015; Paolo et al., 2015; Pritchard et al., 2009; Wuite et al., 2015). Therefore, the detailed long-term observations of the dynamic response of the former ice shelf tributaries allows us to better understand the effects of ice shelf disintegration on the dynamics and mass balances of ice sheets, in order to test and improve projections and models (e.g. Rack et al., 2000). The results from this work have already been employed to model the impact of ice shelf disintegration on the tributary glaciers (Royston and Gudmundsson, 2016). Along the northern Antarctic Peninsula, the glaciers on the east and west side showed diverse temporal trends in ice dynamics and mass balances. On the east coast, all glaciers retreated and a general deceleration of the glaciers north of the former Prince-Gustav Ice Shelf was found, whereas nearly all glaciers affected by ice shelf collapse showed a similar evolution as discussed above. The non-uniform ice dynamics on the west side could be associated with the geometries of the individual glacier basin. Negative mass balances on the east side are compensated by the positive mass balances on the west side. The high accumulation rates along the west coast overbalance the increasing ice discharge and reported surface lowering at the termini, as assumed by Kunz et al. (2012). Consequently, the recent imbalance estimations for the whole study site resulted in balanced or even positive mass budgets, depending on the applied ice thickness and climatic mass balance conditions. These findings are contrary to results from previous studies, 93 Conclusions and Outlook which obtained negative mass balances, but for different periods (e.g. Scambos et al., 2014 for the interval 2003-2008). However, isostatic uplift measurements in this region support the balanced or positive mass balance estimations (Matt King, personal communication). Moreover, the observed cooling trend at the Antarctic Peninsula since the beginning of the 21st century (Turner et al., 2016) as well as the significant reduction of mass loss at the former ice shelf tributaries further support the obtained findings. All studies suggest that the major sources of uncertainties in mass balance computations are the ice-thickness and climatic mass balance. In-situ measurements are quite rare at the Antarctic Peninsula due to the complex topography and the harsh climatic conditions. The latest modeling results provide the best available regional estimates of ice thickness and climatic mass balances (e.g. Huss and Farinotti, 2014; van Wessem et al., 2016). However, the large spatial variability of climatic parameters (e.g. Turner, 2002) and steep topographic gradients (e.g. Cook et al., 2012) limit the quality of the output of the model (e.g. van Wessem et al., 2016). Hence, more field observations for calibration and validation as well as the integration of new findings and techniques will help to further improve the models. The changing global climatic system influences the strongly spatially and temporally variable regional climate of the Antarctic Peninsula Ice Sheet (e.g. Turner et al., 2016). The observed weakening of the ice shelves (e.g. Holland et al., 2015; Wuite et al., 2015) and the strong recession of glaciers on the south western Antarctic Peninsula, which is associated with mid-ocean warming (Cook et al., 2016), might lead to further dramatic mass loss events, like the ones investigated at the north-eastern Antarctic Peninsula. Consequently, continuing observations of the Antarctic Peninsula Ice Sheet are needed to identify potential risks and to monitor the ongoing changes. The obtained data will provide important information to global change analyses, such as SLR projections, on which policy maker refer. Moreover, the presented studies prove the potential of various remote sensing methods to study glaciological processes in rapidly changing regions like the Antarctic Peninsula, but also address limiting factors and error sources. New remote sensing missions have been started recently or will be launched soon. They will allow us to monitor the ongoing changes on Earth at unprecedented temporal density and quality.

94 Acknowledgements

11 Acknowledgements

First of all I want to thank Prof. Dr. Matthias Braun, who initiated this PhD project. He organized the financial support and access to the large amount of remote sensing data. Moreover, I want to thank him for all his support during the work, interesting, motivating and fruitful discussions, fairness, always having a good mood, the possibility to attend numerous conferences, workshops and field trips and joint climbing action. I also want to thank Dr. Wolfgang Rack for showing interest in my work and being a referee of my PhD thesis. I also want to thank all members of the Institute of Geography, Friedrich-Alexander University Erlangen-Nuremberg, for interesting discussions, administrative support and time spent together. Special thanks goes to everyone from the GIS & Remote Sensing working group for the constructive and efficient collaboration, the nice working environment, helpfulness and lunch breaks in our convenience lounge “Sofa-Ecke”. Further, I need to thank Melanie Rankl for helping me to get started in the field of glaciology and remote sensing, the interesting discussions and for sharing the office (sorry, for the disturbances during field campaign preparations) as well as Saurabh Vijay for introducing me in the processing of TanDEM-X data and tasty Indian food. Moreover, I want to thank all project collaborators and co-authors of the studies, especially Sebastián Marinsek from Instituto Antártico Argentino for logistical and personal support of the field campaigns. I also want to acknowledge the logistic support by Alfred Wegener Institut (AWI), Dirección Nacional del Antártico (DNA) as well as the Marambio station staff including the helicopter and Twin Otter crews. Additionally, I want to thank Prof. Dr. Jorge Arigony-Neto for hosting my research stay at the Univeridade Federal do Rio Grande (FURG) and Juliana Costi for helping me to relocate in Brazil. This work was supported by the Deutsche Forschungsgemeinschaft (DFG) within in the framework of the priority programme “Antarctic Research with comparative investigations in Arctic ice areas”; Grant number: BR 2105/9-1. Additional support was granted by the Helmholtz-Alliance “Remote Sensing of Earth System Dynamics” and European Commission 7th Framework Programme International Research Staff Exchange Scheme IMCONet project (FP7 IRSES, action no. 319718) and the Friedrich-Alexander University of Erlangen-Nuremberg. Remote sensing data was kindly provided by Deutsches Zentrum für Luft und Raumfahr (DLR), European Space Agency (ESA), National Aeronautics and Space Administration (NASA), Japan Aerospace Exploration Agency (JAXA), Alaska Satellite Facility (ASF), United States Geological Survey (USGS), National Snow and Ice Data Center (NSICDC), SPOT 5 stereoscopic survey of Polar Ice: Reference Images and Topographies (SPIRIT) project. Last but not least, I want to thank all of my friends and family members, particularly my mother. Special thanks goes to my girlfriend Theresa Bretz for supporting me for many years and the good times and experiences we have had and will have together.

95 References

12 References

Abdalati, W., Zwally, H.J., Bindschadler, R., Csatho, B., Farrell, S.L., Fricker, H.A., Harding, D., Kwok, R., Lefsky, M., Markus, T., Marshak, A., Neumann, T., Palm, S., Schutz, B., Smith, B., Spinhirne, J., Webb, C., 2010. The ICESat-2 Laser Altimetry Mission. Proc. IEEE 98, 735–751. doi:10.1109/JPROC.2009.2034765 Amundson, J.M., Fahnestock, M., Truffer, M., Brown, J., Lüthi, M.P., Motyka, R.J., 2010. Ice mélange dynamics and implications for terminus stability, Jakobshavn Isbræ, Greenland. J. Geophys. Res. Earth Surf. 115, F01005. doi:10.1029/2009JF001405 Atwood, D.K., Meyer, F., Arendt, A., 2010. Using L-band SAR coherence to delineate glacier extent. Can. J. Remote Sens. 36, S186–S195. doi:10.5589/m10-014 Barrand, N.E., Hindmarsh, R.C.A., Arthern, R.J., Williams, C.R., Mouginot, J., Scheuchl, B., Rignot, E., Ligtenberg, S.R.M., Van Den Broeke, M.R., Edwards, T.L., Cook, A.J., Simonsen, S.B., 2013a. Computing the volume response of the Antarctic Peninsula ice sheet to warming scenarios to 2200. J. Glaciol. 59, 397–409. doi:10.3189/2013JoG12J139 Barrand, N.E., Vaughan, D.G., Steiner, N., Tedesco, M., Kuipers Munneke, P., van den Broeke, M.R., Hosking, J.S., 2013b. Trends in Antarctic Peninsula surface melting conditions from observations and regional climate modeling. J. Geophys. Res. Earth Surf. 118, 315–330. doi:10.1029/2012JF002559 Benn, D.I., Evans, D.J.A., 1998. Glaciers & Glaciation. Arnold. Benn, D.I., Hulton, N.R.J., Mottram, R.H., 2007. “Calving laws”, “sliding laws” and the stability of tidewater glaciers. Ann. Glaciol. 46, 123–130. doi:10.3189/172756407782871161 Berthier, E., Scambos, T.A., Shuman, C.A., 2012. Mass loss of Larsen B tributary glaciers (Antarctic Peninsula) unabated since 2002. Geophys. Res. Lett. 39, L13501. doi:10.1029/2012GL051755 Berthier, E., Schiefer, E., Clarke, G.K.C., Menounos, B., Rémy, F., 2010. Contribution of Alaskan glaciers to sea-level rise derived from satellite imagery. Nat. Geosci. 3, 92–95. doi:10.1038/ngeo737 Bolch, T., Sandberg Sørensen, L., Simonsen, S.B., Mölg, N., Machguth, H., Rastner, P., Paul, F., 2013. Mass loss of Greenland’s glaciers and ice caps 2003–2008 revealed from ICESat laser altimetry data. Geophys. Res. Lett. 40, 875–881. doi:10.1002/grl.50270 Braun, M., Humbert, A., 2009. Recent Retreat of Wilkins Ice Shelf Reveals New Insights in Ice Shelf Breakup Mechanisms. IEEE Geosci. Remote Sens. Lett. 6, 263–267. doi:10.1109/LGRS.2008.2011925 Braun, M., Humbert, A., Moll, A., 2009. Changes of Wilkins Ice Shelf over the past 15 years and inferences on its stability. The Cryosphere 3, 41–56. doi:10.5194/tc-3-41-2009 Brenner, A.C., DiMarzio, J.P., Zwally, H.J., 2007. Precision and Accuracy of Satellite Radar and Laser Altimeter Data Over the Continental Ice Sheets. IEEE Trans. Geosci. Remote Sens. 45, 321–331. doi:10.1109/TGRS.2006.887172 Burgess, E.W., Forster, R.R., Larsen, C.F., Braun, M., 2012. Surge dynamics on Bering Glacier, Alaska, in 2008–2011. The Cryosphere 6, 1251–1262. doi:10.5194/tc-6-1251-2012 Burton-Johnson, A., Black, M., Fretwell, P.T., Kaluza-Gilbert, J., 2016. A fully automated methodology for differentiating rock from snow, clouds and sea in Antarctica from Landsat imagery: A new rock outcrop map and area estimation for the entire Antarctic continent. Cryosphere Discuss. 1–16. doi:10.5194/tc-2016-56 Cape, M.R., Vernet, M., Skvarca, P., Marinsek, S., Scambos, T., Domack, E., 2015. Foehn winds link climate-driven warming to ice shelf evolution in Antarctica. J. Geophys. Res. Atmospheres 2015JD023465. doi:10.1002/2015JD023465

96 References Cogley, J.G., Hock, R., Rasmussen, L.A., Arendt, A.A., Bauder, A., Braithwaite, R.J., Jansson, P., Kaser, G., Möller, M., Nicholson, L., Zemp, M., 2011. Glossary of Glacier Mass Balance and Related Terms. IHP-VII Technical Documents in Hydrology No. 86, IACS Contribution No. 2, UNESCO-IHP, Paris. Polar Rec. 48. doi:10.1017/S0032247411000805 Cook, A. j., Vaughan, D. g., Luckman, A. j., Murray, T., 2014. A new Antarctic Peninsula glacier basin inventory and observed area changes since the 1940s. Antarct. Sci. 26, 614–624. doi:10.1017/S0954102014000200 Cook, A.J., Fox, A.J., Vaughan, D.G., Ferrigno, J.G., 2005. Retreating Glacier Fronts on the Antarctic Peninsula over the Past Half-Century. Science 308, 541–544. doi:10.1126/science.1104235 Cook, A.J., Holland, P.R., Meredith, M.P., Murray, T., Luckman, A., Vaughan, D.G., 2016. Ocean forcing of glacier retreat in the western Antarctic Peninsula. Science 353, 283–286. doi:10.1126/science.aae0017 Cook, A.J., Murray, T., Luckman, A., Vaughan, D.G., Barrand, N.E., 2012. A new 100-m Digital Elevation Model of the Antarctic Peninsula derived from ASTER Global DEM: methods and accuracy assessment. Earth Syst. Sci. Data 4, 129–142. doi:10.5194/essd-4-129-2012 Cook, A.J., Vaughan, D.G., 2010. Overview of areal changes of the ice shelves on the Antarctic Peninsula over the past 50 years. The Cryosphere 4, 77–98. doi:10.5194/tc-4-77-2010 Cooper, A. p. r., 1997. Historical observations of Prince Gustav Ice Shelf. Polar Rec. 33, 285–294. doi:10.1017/S0032247400025389 Corr, H.F.J., Doake, C.S.M., Jenkins, A., Vaughan, D.G., 2001. Investigations of an “ice plain” in the mouth of Pine Island Glacier, Antarctica. J. Glaciol. 47, 51–57. doi:10.3189/172756501781832395 Davies, B.J., Carrivick, J.L., Glasser, N.F., Hambrey, M.J., Smellie, J.L., 2012. Variable glacier response to atmospheric warming, northern Antarctic Peninsula, 1988–2009. The Cryosphere 6, 1031–1048. doi:10.5194/tc-6-1031-2012 De Angelis, H., Skvarca, P., 2003. Glacier Surge After Ice Shelf Collapse. Science 299, 1560– 1562. doi:10.1126/science.1077987 de Lange, R., Luckman, A., Murray, T., 2007. Improvement of Satellite Radar Feature Tracking for Ice Velocity Derivation by Spatial Frequency Filtering. IEEE Trans. Geosci. Remote Sens. 45, 2309–2318. doi:10.1109/TGRS.2007.896615 Dee, D.P., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U., Balmaseda, M.A., Balsamo, G., Bauer, P., Bechtold, P., Beljaars, A.C.M., van de Berg, L., Bidlot, J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Geer, A.J., Haimberger, L., Healy, S.B., Hersbach, H., Hólm, E.V., Isaksen, L., Kållberg, P., Köhler, M., Matricardi, M., McNally, A.P., Monge-Sanz, B.M., Morcrette, J.-J., Park, B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., Vitart, F., 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597. doi:10.1002/qj.828 DLR-IMF, Balss, U., Rossi, C., Fritz, T., Balzer, W., Breit, H., Duque, S., 2012. TanDEM-X CoSSC Generation and Interferometric Considerations. Doake, C.S.M., Vaughan, D.G., 1991. Rapid disintegration of the Wordie Ice Shelf in response to atmospheric warming. Nature 350, 328–330. doi:10.1038/350328a0 Dutrieux, P., Rydt, J.D., Jenkins, A., Holland, P.R., Ha, H.K., Lee, S.H., Steig, E.J., Ding, Q., Abrahamsen, E.P., Schröder, M., 2014. Strong Sensitivity of Pine Island Ice-Shelf Melting to Climatic Variability. Science 343, 174–178. doi:10.1126/science.1244341 Favier, V., Agosta, C., Parouty, S., Durand, G., Delaygue, G., Gallée, H., Drouet, A.-S., Trouvilliez, A., Krinner, G., 2013. An updated and quality controlled surface mass balance dataset for Antarctica. The Cryosphere 7, 583–597. doi:10.5194/tc-7-583-2013

97 References Fernandoy, F., Meyer, H., Tonelli, M., 2012. Stable water isotopes of precipitation and firn cores from the northern Antarctic Peninsula region as a proxy for climate reconstruction. The Cryosphere 6, 313–330. doi:10.5194/tc-6-313-2012 Ferrigno, J.G., Cook, A.J., Foley, K.M., Williams, R.S., Swithinbank, C., Fox, A.J., Thomson, J.W., Sievers, J., 2006. Coastal-change and glaciological map of the Trinity Peninsula area and South Shetland Islands, Antarctica, 1843-2001. Flament, T., Rémy, F., 2012. Dynamic thinning of Antarctic glaciers from along-track repeat radar altimetry. J. Glaciol. 58, 830–840. doi:10.3189/2012JoG11J118 Fletcher, K., European Space Agency (Eds.), 2007. InSAR principles: guidelines for SAR interferometry processing and interpretation, ESA TM. ESA Publications, ESTEC, Noordwijk, the Netherlands. Fountain, A.G., Krimmel, R.M., Trabant, D., Geological Survey (U.S.), 1997. A strategy for monitoring glaciers. U.S. G.P.O. ; Free on applications to the U.S. Geological Survey, Information Services, [Washington]; Denver, CO. Fretwell, P., Pritchard, H.D., Vaughan, D.G., Bamber, J.L., Barrand, N.E., Bell, R., Bianchi, C., Bingham, R.G., Blankenship, D.D., Casassa, G., Catania, G., Callens, D., Conway, H., Cook, A.J., Corr, H.F.J., Damaske, D., Damm, V., Ferraccioli, F., Forsberg, R., Fujita, S., Gim, Y., Gogineni, P., Griggs, J.A., Hindmarsh, R.C.A., Holmlund, P., Holt, J.W., Jacobel, R.W., Jenkins, A., Jokat, W., Jordan, T., King, E.C., Kohler, J., Krabill, W., Riger-Kusk, M., Langley, K.A., Leitchenkov, G., Leuschen, C., Luyendyk, B.P., Matsuoka, K., Mouginot, J., Nitsche, F.O., Nogi, Y., Nost, O.A., Popov, S.V., Rignot, E., Rippin, D.M., Rivera, A., Roberts, J., Ross, N., Siegert, M.J., Smith, A.M., Steinhage, D., Studinger, M., Sun, B., Tinto, B.K., Welch, B.C., Wilson, D., Young, D.A., Xiangbin, C., Zirizzotti, A., 2013. Bedmap2: improved ice bed, surface and thickness datasets for Antarctica. The Cryosphere 7, 375–393. doi:10.5194/tc-7-375-2013 Fritz, T., Rossi, C., Yague-Martinez, N., Rodriguez-Gonzalez, F., Lachaise, M., Breit, H., 2011. Interferometric processing of TanDEM-X data. IEEE, pp. 2428–2431. doi:10.1109/IGARSS.2011.6049701 Fürst, J.J., Durand, G., Gillet-Chaulet, F., Tavard, L., Rankl, M., Braun, M., Gagliardini, O., 2016. The safety band of Antarctic ice shelves. Nat. Clim. Change 6, 479–482. doi:10.1038/nclimate2912 Gardner, A.S., Moholdt, G., Cogley, J.G., Wouters, B., Arendt, A.A., Wahr, J., Berthier, E., Hock, R., Pfeffer, W.T., Kaser, G., Ligtenberg, S.R.M., Bolch, T., Sharp, M.J., Hagen, J.O., Broeke, M.R. van den, Paul, F., 2013. A Reconciled Estimate of Glacier Contributions to Sea Level Rise: 2003 to 2009. Science 340, 852–857. doi:10.1126/science.1234532 Glasser, N.F., Scambos, T.A., 2008. A structural glaciological analysis of the 2002 Larsen B ice- shelf collapse. J. Glaciol. 54, 3–16. Glasser, N.F., Scambos, T.A., Bohlander, J., Truffer, M., Pettit, E., Davies, B.J., 2011. From ice- shelf tributary to tidewater glacier: continued rapid recession, acceleration and thinning of Röhss Glacier following the 1995 collapse of the Prince Gustav Ice Shelf, Antarctic Peninsula. J. Glaciol. 57, 397–406. Gleick, P.H., Pacific Institute for Studies in Development, Environment, and Security, Stockholm Environment Institute (Eds.), 1993. Water in crisis: a guide to the world’s fresh water resources. Oxford University Press, New York. Goldberg, D., Holland, D.M., Schoof, C., 2009. Grounding line movement and ice shelf buttressing in marine ice sheets. J. Geophys. Res. Earth Surf. 114, F04026. doi:10.1029/2008JF001227 Goodwin, B.P., 2013. Recent Environmental Changes on the Antarctic Peninsula as Recorded in an ice core from the Bruce Plateau. The Ohio State University.

98 References Gray, A.L., Mattar, K.E., Vachon, P.W., 1998. InSAR Results from the RADARSAT Antarctic Mapping Mission Data: Estimation of Glacier Motion Using a Simple Registration Procedure, in: Proceedings of IGARSS. Seattle. Groh, A., Horwath, M., 2016. The method of tailored sensitivity kernels for GRACE mass change estimates. Geophys. Res. Abstr. 18. Haakensen, N., 1986. Glacier Mapping to Confirm Results from Mass-Balance Measurements. Ann. Glaciol. 8, 73–77. doi:10.3198/1986AoG8-1-73-77 Haug, T., Kääb, A., Skvarca, P., 2010. Monitoring ice shelf velocities from repeat MODIS and Landsat data – a method study on the Larsen C ice shelf, Antarctic Peninsula, and 10 other ice shelves around Antarctica. The Cryosphere 4, 161–178. doi:10.5194/tc-4-161-2010 Hellmer, H.H., Kauker, F., Timmermann, R., Determann, J., Rae, J., 2012. Twenty-first-century warming of a large Antarctic ice-shelf cavity by a redirected coastal current. Nature 485, 225–228. doi:10.1038/nature11064 Helm, V., Humbert, A., Miller, H., 2014. Elevation and elevation change of Greenland and Antarctica derived from CryoSat-2. The Cryosphere 8, 1539–1559. doi:10.5194/tc-8-1539- 2014 Herron, M.M., Langway, C.C., 1980. Firn Densification: An Empirical Model. J. Glaciol. 25, 373– 385. doi:10.3198/1980JoG25-93-373-385 Holland, P.R., Brisbourne, A., Corr, H.F.J., McGrath, D., Purdon, K., Paden, J., Fricker, H.A., Paolo, F.S., Fleming, A.H., 2015. Oceanic and atmospheric forcing of Larsen C Ice-Shelf thinning. The Cryosphere 9, 1005–1024. doi:10.5194/tc-9-1005-2015 Holland, P.R., Jenkins, A., Holland, D.M., 2010. Ice and ocean processes in the Bellingshausen Sea, Antarctica. J. Geophys. Res. Oceans 115, C05020. doi:10.1029/2008JC005219 Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., Linden, P.J. van der, Dai, X., Maskell, K., Johnson, C.A. (Eds.), 2001. Climate Change 2001: The Scientific Basis: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge ; New York. Hughes, T.J., 1981. The weak underbelly of the West Antarctic ice sheet. J. Glaciol. 27, 518–525. Hulbe, C.L., MacAyeal, D.R., Denton, G.H., Kleman, J., Lowell, T.V., 2004. Catastrophic ice shelf breakup as the source of Heinrich event icebergs. Paleoceanography 19, PA1004. doi:10.1029/2003PA000890 Huss, M., 2013. Density assumptions for converting geodetic glacier volume change to mass change. The Cryosphere 7, 877–887. doi:10.5194/tc-7-877-2013 Huss, M., Farinotti, D., 2014. A high-resolution bedrock map for the Antarctic Peninsula. The Cryosphere 8, 1261–1273. doi:10.5194/tc-8-1261-2014 IPCC, 2014. Climate Change 2013 - The Physical Science Basis: Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge. IPCC, 2007. Climate Change 2007 - The Physical Science Basis [WWW Document]. Camb. Univ. Press. URL http://www.cambridge.org/de/academic/subjects/earth-and-environmental- science/climatology-and-climate-change/climate-change-2007-physical-science-basis- working-group-i-contribution-fourth-assessment-report-ipcc (accessed 7.12.16). Ivins, E.R., Watkins, M.M., Yuan, D.-N., Dietrich, R., Casassa, G., Rülke, A., 2011. On-land ice loss and glacial isostatic adjustment at the Drake Passage: 2003–2009. J. Geophys. Res. Solid Earth 116, B02403. doi:10.1029/2010JB007607 Jacobs, S.S., Comiso, J.C., 1997. Climate Variability in the Amundsen and Bellingshausen Seas. J. Clim. 10, 697–709. doi:10.1175/1520-0442(1997)010<0697:CVITAA>2.0.CO;2

99 References Jacobs, S.S., Jenkins, A., Giulivi, C.F., Dutrieux, P., 2011. Stronger ocean circulation and increased melting under Pine Island Glacier ice shelf. Nat. Geosci. 4, 519–523. doi:10.1038/ngeo1188 Jansen, D., Luckman, A.J., Cook, A., Bevan, S., Kulessa, B., Hubbard, B., Holland, P.R., 2015. Brief Communication: Newly developing rift in Larsen C Ice Shelf presents significant risk to stability. The Cryosphere 9, 1223–1227. doi:10.5194/tc-9-1223-2015 Jensen, J.R., 2007. Remote sensing of the environment: an earth resource perspective, 2nd ed. ed, Prentice Hall series in geographic information science. Pearson Prentice Hall, Upper Saddle River, NJ. Jiskoot, H., Curran, C.J., Tessler, D.L., Shenton, L.R., 2009. Changes in Clemenceau Icefield and Chaba Group glaciers, Canada, related to hypsometry, tributary detachment, length–slope and area–aspect relations. Ann. Glaciol. 50, 133–143. doi:10.3189/172756410790595796 Khazendar, A., Borstad, C.P., Scheuchl, B., Rignot, E., Seroussi, H., 2015. The evolving instability of the remnant Larsen B Ice Shelf and its tributary glaciers. Earth Planet. Sci. Lett. 419, 199–210. doi:10.1016/j.epsl.2015.03.014 Khazendar, A., Rignot, E., Larour, E., 2011. Acceleration and spatial rheology of Larsen C Ice Shelf, Antarctic Peninsula: ACCELERATION AND RHEOLOGY OF LARSEN C. Geophys. Res. Lett. 38, n/a-n/a. doi:10.1029/2011GL046775 King, M.A., Bingham, R.J., Moore, P., Whitehouse, P.L., Bentley, M.J., Milne, G.A., 2012. Lower satellite-gravimetry estimates of Antarctic sea-level contribution. Nature 491, 586–589. doi:10.1038/nature11621 Korona, J., Berthier, E., Bernard, M., Rémy, F., Thouvenot, E., 2009. SPIRIT. SPOT 5 stereoscopic survey of polar ice: reference images and topographies during the fourth International Polar Year (2007–2009). ISPRS J. Photogramm. Remote Sens. 64, 204–212. Krabill, W.B., 2010. updated 2014. IceBridge ATM L2 Icessn Elevation, Slope, and Roughness. Version 2, [from 2011-11-14]. Boulder, Colorado USA: NASA DAAC at the National Snow and Ice Data Center. [WWW Document]. URL http://nsidc.org/data/ilatm2.html. Krimmel, R.M., 1989. Mass Balance and Volume of South Cascade Glacier, Washington 1958– 1985, in: Oerlemans, J. (Ed.), Glacier Fluctuations and Climatic Change, Glaciology and Quaternary Geology. Springer Netherlands, pp. 193–206. Kuipers Munneke, P., Picard, G., van den Broeke, M.R., Lenaerts, J.T.M., van Meijgaard, E., 2012. Insignificant change in Antarctic snowmelt volume since 1979. Geophys. Res. Lett. 39, L01501. doi:10.1029/2011GL050207 Kunz, M., King, M.A., Mills, J.P., Miller, P.E., Fox, A.J., Vaughan, D.G., Marsh, S.H., 2012. Multi- decadal glacier surface lowering in the Antarctic Peninsula. Geophys. Res. Lett. 39, L19502. doi:10.1029/2012GL052823 Lenaerts, J.T.M., van den Broeke, M.R., van de Berg, W.J., van Meijgaard, E., Kuipers Munneke, P., 2012. A new, high-resolution surface mass balance map of Antarctica (1979–2010) based on regional atmospheric climate modeling. Geophys. Res. Lett. 39, L04501. doi:10.1029/2011GL050713 Leuschen, C., Prasad Gogineni, Fernando Rodriguez, John Paden, Chris Allen, 2010. updated 2014. IceBridge MCoRDS L2 Ice Thickness, [from 2011-11-14]. Boulder, Colorado USA: National Snow and Ice Data Center. [WWW Document]. URL http://nsidc.org/data/irmcr2.html. Li, J., Zwally, H.J., 2011. Modeling of firn compaction for estimating ice-sheet mass change from observed ice-sheet elevation change. Ann. Glaciol. 52, 1–7. doi:10.3189/172756411799096321 Love, A., 1985. In memory of Carl A. Wiley. IEEE Antennas Propag. Soc. Newsl. 27, 17–18. doi:10.1109/MAP.1985.27810

100 References Luckman, A., Elvidge, A., Jansen, D., Kulessa, B., Kuipers Munneke, P., King, J., Barrand, N.E., 2014. Surface melt and ponding on Larsen C Ice Shelf and the impact of föhn winds. Antarct. Sci. 26, 625–635. doi:10.1017/S0954102014000339 Marshall, G.J., Orr, A., van Lipzig, N.P.M., King, J.C., 2006. The Impact of a Changing Southern Hemisphere Annular Mode on Antarctic Peninsula Summer Temperatures. J. Clim. 19, 5388–5404. doi:10.1175/JCLI3844.1 Martin, P.J., Peel, D.A., 1978. The Spatial Distribution of 10 m Temperatures in the Antarctic Peninsula. J. Glaciol. 20, 311–317. doi:10.3198/1978JoG20-83-311-317 Massom, R., Lubin, D., 2006. Polar remote sensing Vol. II: Ice Sheets. Springer [u.a., Berlin [u.a. McMillan, M., Shepherd, A., Sundal, A., Briggs, K., Muir, A., Ridout, A., Hogg, A., Wingham, D., 2014. Increased ice losses from Antarctica detected by CryoSat-2. Geophys. Res. Lett. 41, 2014GL060111. doi:10.1002/2014GL060111 McNabb, R.W., Hock, R., O’Neel, S., Rasmussen, L.A., Ahn, Y., Braun, M., Conway, H., Herreid, S., Joughin, I., Pfeffer, W.T., Smith, B.E., Truffer, M., 2012. Using surface velocities to calculate ice thickness and bed topography: a case study at Columbia Glacier, Alaska, USA. J. Glaciol. 58, 1151–1164. doi:10.3189/2012JoG11J249 Meier, M.F., Post, A., 1987. Fast tidewater glaciers. J. Geophys. Res. Solid Earth 92, 9051–9058. doi:10.1029/JB092iB09p09051 Meredith, M.P., King, J.C., 2005. Rapid climate change in the ocean west of the Antarctic Peninsula during the second half of the 20th century. Geophys. Res. Lett. 32, L19604. doi:10.1029/2005GL024042 Moon, T., Joughin, I., Smith, B., 2015. Seasonal to multiyear variability of glacier surface velocity, terminus position, and sea ice/ice mélange in northwest Greenland. J. Geophys. Res. Earth Surf. 120, 2015JF003494. doi:10.1002/2015JF003494 Moon, T., Joughin, I., Smith, B., van den Broeke, M.R., van de Berg, W.J., Noël, B., Usher, M., 2014. Distinct patterns of seasonal Greenland glacier velocity. Geophys. Res. Lett. 41, 2014GL061836. doi:10.1002/2014GL061836 Morris, E.M., Vaughan, D.G., 2003. Spatial and Temporal Variation of Surface Temperature on the Antarctic Peninsula And The Limit of Viability of Ice Shelves, in: Domack, E., Levente, A., Burnet, A., Bindschadler, R., Convey, P., Kirby, M. (Eds.), Antarctic Peninsula Climate Variability: Historical and Paleoenvironmental Perspectives. American Geophysical Union, pp. 61–68. NASA LP DAAC, 2007. ASTER On-Demand L3 DEM and Orthorectified Images, GeoTIF Format. Nield, G.A., Barletta, V.R., Bordoni, A., King, M.A., Whitehouse, P.L., Clarke, P.J., Domack, E., Scambos, T.A., Berthier, E., 2014. Rapid bedrock uplift in the Antarctic Peninsula explained by viscoelastic response to recent ice unloading. Earth Planet. Sci. Lett. 397, 32–41. doi:10.1016/j.epsl.2014.04.019 Orr, A., Cresswell, D., Marshall, G.J., Hunt, J.C.R., Sommeria, J., Wang, C.G., Light, M., 2004. A “low-level” explanation for the recent large warming trend over the western Antarctic Peninsula involving blocked winds and changes in zonal circulation. Geophys. Res. Lett. 31, L06204. doi:10.1029/2003GL019160 Osmanoglu, B., Braun, M., Hock, R., Navarro, F.J., 2013. Surface velocity and ice discharge of the ice cap on King George Island, Antarctica. Ann. Glaciol. 54, 111–119. doi:10.3189/2013AoG63A517 Osmanoglu, B., Navarro, F.J., Hock, R., Braun, M., Corcuera, M.I., 2014. Surface velocity and mass balance of Livingston Island ice cap, Antarctica. The Cryosphere 8, 1807–1823. doi:10.5194/tc-8-1807-2014

101 References Padman, L., Fricker, H.A., Coleman, R., Howard, S., Erofeeva, L., 2002. A new tide model for the Antarctic ice shelves and seas. Ann. Glaciol. 34, 247–254. doi:10.3189/172756402781817752 Paolo, F.S., Fricker, H.A., Padman, L., 2015. Volume loss from Antarctic ice shelves is accelerating. Science 348, 327–331. doi:10.1126/science.aaa0940 Paterson, W.S.B., 2000. Physics of Glaciers, Third Edition, 3 edition. ed. Butterworth-Heinemann, Oxford ; Woburn, MA. Pearce, F., 2007. With speed and violence: why scientists fear tipping points in climate change. Beacon Press, Boston. Peel, D.A., 1992. Spatial temperature and accumulation rate variations at the Antarctic Peninsula, in: The Contributions of Antarctic Peninsula to Sea Level Rise, CEC Project Report EPOC- CT90-0015. pp. 11–15. Pellikka, P.K.E., Rees, G. (Eds.), 2010. Remote sensing of glaciers: techniques for topographic, spatial, and thematic mapping of glaciers. CRC Press, Boco Raton [Fla.]. Polar Space Task Group, 2016. Data Compendium - Summary Documentation of SAR Satellite Data Collections, Plans and Activities. Presented at the olar Space Task Group - SAR Coordination Working Group, Saint Hubert, QC, Canada, p. 44. Pope, A., Rees, W., Fox, A., Fleming, A., 2014. Open Access Data in Polar and Cryospheric Remote Sensing. Remote Sens. 6, 6183–6220. doi:10.3390/rs6076183 Potocki, M., Mayewski, P.A., Kurbatov, A., Handley, M., Simoes, J.C., Jaña, R., 2011. Detailed Glaciochemical Records from a Northern Antarctic Peninsula Site - . AGU Fall Meet. Abstr. 43. Pritchard, H., 2005. Glacier surge dynamics of Sortebræ, east Greenland, from synthetic aperture radar feature tracking. J. Geophys. Res. 110. doi:10.1029/2004JF000233 Pritchard, H.D., Arthern, R.J., Vaughan, D.G., Edwards, L.A., 2009. Extensive dynamic thinning on the margins of the Greenland and Antarctic ice sheets. Nature 461, 971–975. doi:10.1038/nature08471 Pritchard, H.D., Ligtenberg, S.R.M., Fricker, H.A., Vaughan, D.G., van den Broeke, M.R., Padman, L., 2012. Antarctic ice-sheet loss driven by basal melting of ice shelves. Nature 484, 502– 505. doi:10.1038/nature10968 Pritchard, H.D., Vaughan, D.G., 2007. Widespread acceleration of tidewater glaciers on the Antarctic Peninsula. J. Geophys. Res. 112. doi:10.1029/2006JF000597 Pudsey, C.J., Evans, J., Domack, E.W., Morris, P., Valle, R.A.D., 2001. Bathymetry and acoustic facies beneath the former Larsen-A and Prince Gustav ice shelves, north-west Weddell Sea. Antarct. Sci. 13, 312–322. doi:10.1017/S095410200100044X Rack, W., Doake, C.S., Rott, H., Siegel, A., Skvarca, P., 2000. Interferometric analysis of the deformation pattern of the northern Larsen Ice Shelf, Antarctic Peninsula, compared to field measurements and numerical modeling. Ann. Glaciol. 31, 205–210. Rack, W., Rott, H., 2004. Pattern of retreat and disintegration of the Larsen B ice shelf, Antarctic Peninsula. Ann. Glaciol. 39, 505–510. Rack, W., Rott, H., 2003. Further retreat of the northern Larsen Ice Shelf and collapse of Larsen B, in: 16th International Workshop of the Forum for Research on Ice Shelf Processes (FRISP), June 25. Rankl, M., Braun, M., 2016. Glacier elevation and mass changes over the central Karakoram region estimated from TanDEM-X and SRTM/X-SAR digital elevation models. Ann. Glaciol. 51. Rankl, M., Kienholz, C., Braun, M., 2014. Glacier changes in the Karakoram region mapped by multimission satellite imagery. The Cryosphere 8, 977–989. doi:10.5194/tc-8-977-2014

102 References Rau, F., Braun, M., 2002. The regional distribution of the dry-snow zone on the Antarctic Peninsula north of 70°S. Ann. Glaciol. 34, 95–100. doi:10.3189/172756402781817914 Rees, G., 2006. Remote sensing of snow and ice. Taylor & Francis, Boca Raton. Riffenburgh, B., 2007. Encyclopedia of the Antarctic. Taylor & Francis. Rignot, E., 2008. Changes in West Antarctic ice stream dynamics observed with ALOS PALSAR data. Geophys. Res. Lett. 35, L12505. doi:10.1029/2008GL033365 Rignot, E., Casassa, G., Gogineni, P., Krabill, W., Rivera, A., Thomas, R., 2004. Accelerated ice discharge from the Antarctic Peninsula following the collapse of Larsen B ice shelf. Geophys. Res. Lett. 31. doi:10.1029/2004GL020697 Rignot, E., Casassa, G., Gogineni, S., Kanagaratnam, P., Krabill, W., Pritchard, H., Rivera, A., Thomas, R., Turner, J., Vaughan, D., 2005. Recent ice loss from the Fleming and other glaciers, Wordie Bay, West Antarctic Peninsula. Geophys. Res. Lett. 32, L07502. doi:10.1029/2004GL021947 Rignot, E., Echelmeyer, K., Krabill, W., 2001. Penetration depth of interferometric synthetic- aperture radar signals in snow and ice. Geophys. Res. Lett. 28, 3501–3504. Rignot, E., Jacobs, S., Mouginot, J., Scheuchl, B., 2013. Ice-Shelf Melting Around Antarctica. Science 341, 266–270. doi:10.1126/science.1235798 Rignot, E., Mouginot, J., Scheuchl, B., 2011a. Ice Flow of the Antarctic Ice Sheet. Science 333, 1427–1430. doi:10.1126/science.1208336 Rignot, E., Mouginot, J., Scheuchl, B., 2011b. Antarctic grounding line mapping from differential satellite radar interferometry: GROUNDING LINE OF ANTARCTICA. Geophys. Res. Lett. 38, n/a-n/a. doi:10.1029/2011GL047109 Rosen, P.A., Hensley, S., Joughin, I.R., Li, F.K., Madsen, S.N., Rodriguez, E., Goldstein, R.M., 2000. Synthetic aperture radar interferometry. Proc. IEEE 88, 333–382. Rott, H., Floricioiu, D., Wuite, J., Scheiblauer, S., Nagler, T., Kern, M., 2014. Mass changes of outlet glaciers along the Nordensjköld Coast, northern Antarctic Peninsula, based on TanDEM-X satellite measurements. Geophys. Res. Lett. 2014GL061613. doi:10.1002/2014GL061613 Rott, H., Müller, F., Nagler, T., Floricioiu, D., 2011. The imbalance of glaciers after disintegration of Larsen-B ice shelf, Antarctic Peninsula. The Cryosphere 5, 125–134. doi:10.5194/tc-5-125- 2011 Rott, H., Rack, W., Nagler, T., 2007. Increased export of grounded ice after the collapse of northern Larsen Ice Shelf, Antarctic Peninsula, observed by Envisat ASAR, in: Geoscience and Remote Sensing Symposium, 2007. IGARSS 2007. IEEE International. IEEE, pp. 1174– 1176. Rott, H., Rack, W., Skvarca, P., De Angelis, H., 2002. Northern Larsen ice shelf, Antarctica: further retreat after collapse. Ann. Glaciol. 34, 277–282. Rott, H., Skvarca, P., Nagler, T., 1996. Rapid Collapse of Northern Larsen Ice Shelf, Antarctica. Science 271, 788–792. doi:10.1126/science.271.5250.788 Royston, S., Gudmundsson, G.H., 2016. Changes in ice-shelf buttressing following the collapse of Larsen A Ice Shelf, Antarctica, and the resulting impact on tributaries. J. Glaciol. FirstView, 1–7. doi:10.1017/jog.2016.77 Sasgen, I., Konrad, H., Ivins, E.R., Van den Broeke, M.R., Bamber, J.L., Martinec, Z., Klemann, V., 2013. Antarctic ice-mass balance 2003 to 2012: regional reanalysis of GRACE satellite gravimetry measurements with improved estimate of glacial-isostatic adjustment based on GPS uplift rates. The Cryosphere 7, 1499–1512. doi:10.5194/tc-7-1499-2013

103 References Scambos, T.A., Berthier, E., Haran, T., Shuman, C.A., Cook, A.J., Ligtenberg, S.R.M., Bohlander, J., 2014. Detailed ice loss pattern in the northern Antarctic Peninsula: widespread decline driven by ice front retreats. The Cryosphere 8, 2135–2145. doi:10.5194/tc-8-2135-2014 Scambos, T.A., Bohlander, J., Shuman, C.A., Skvarca, P., 2004. Glacier acceleration and thinning after ice shelf collapse in the Larsen B embayment, Antarctica. Geophys. Res. Lett. 31. doi:10.1029/2004GL020670 Scambos, T.A., Dutkiewicz, M.J., Wilson, J.C., Bindschadler, R.A., 1992. Application of image cross-correlation to the measurement of glacier velocity using satellite image data. Remote Sens. Environ. 42, 177–186. doi:10.1016/0034-4257(92)90101-O Scambos, T.A., Haran, T.M., Fahnestock, M.A., Painter, T.H., Bohlander, J., 2007. MODIS-based Mosaic of Antarctica (MOA) data sets: Continent-wide surface morphology and snow grain size. Remote Sens. Environ. 111, 242–257. doi:10.1016/j.rse.2006.12.020 Scambos, T., Hulbe, C., Fahnestock, M., 2003. Climate-Induced Ice Shelf Disintegration in the Antarctic Peninsula, in: Domack, E., Levente, A., Burnet, A., Bindschadler, R., Convey, P., Kirby, tthew (Eds.), Antarctic Peninsula Climate Variability: Historical and Paleoenvironmental Perspectives. American Geophysical Union, pp. 79–92. Scheuchl, B., Mouginot, J., Rignot, E., 2012. Ice velocity changes in the Ross and Ronne sectors observed using satellite radar data from 1997 and 2009. The Cryosphere 6, 1019–1030. doi:10.5194/tc-6-1019-2012 Schoof, C., 2007. Ice sheet grounding line dynamics: Steady states, stability, and hysteresis. J. Geophys. Res. Earth Surf. 112, F03S28. doi:10.1029/2006JF000664 Schutz, B.E., Zwally, H.J., Shuman, C.A., Hancock, D., DiMarzio, J.P., 2005. Overview of the ICESat Mission. Geophys. Res. Lett. 32, L21S01. doi:10.1029/2005GL024009 Schwerdtfeger, W., 1984. Weather and climate of the Antarctic, Developments in atmospheric science. Elsevier : Distributors for the U.S. and Canada, Elsevier Science Pub. Co, Amsterdam ; New York. Seehaus, T., Marinsek, S., Helm, V., Skvarca, P., Braun, M., 2015. Changes in ice dynamics, elevation and mass discharge of Dinsmoor–Bombardier–Edgeworth glacier system, Antarctic Peninsula. Earth Planet. Sci. Lett. 427, 125–135. doi:10.1016/j.epsl.2015.06.047 Seehaus, T.C., Marinsek, S., Skvarca, P., van Wessem, J.M., Reijmer, C.H., Seco, J.L., Braun, M.H., 2016. Dynamic Response of Sjögren Inlet Glaciers, Antarctic Peninsula, to Ice Shelf Breakup Derived from Multi-Mission Remote Sensing Time Series. Front. Earth Sci. 4. doi:10.3389/feart.2016.00066 Shepherd, A., Ivins, E.R., A, G., Barletta, V.R., Bentley, M.J., Bettadpur, S., Briggs, K.H., Bromwich, D.H., Forsberg, R., Galin, N., Horwath, M., Jacobs, S., Joughin, I., King, M.A., Lenaerts, J.T.M., Li, J., Ligtenberg, S.R.M., Luckman, A., Luthcke, S.B., McMillan, M., Meister, R., Milne, G., Mouginot, J., Muir, A., Nicolas, J.P., Paden, J., Payne, A.J., Pritchard, H., Rignot, E., Rott, H., Sørensen, L.S., Scambos, T.A., Scheuchl, B., Schrama, E.J.O., Smith, B., Sundal, A.V., Angelen, J.H. van, Berg, W.J. van de, Broeke, M.R. van den, Vaughan, D.G., Velicogna, I., Wahr, J., Whitehouse, P.L., Wingham, D.J., Yi, D., Young, D., Zwally, H.J., 2012. A Reconciled Estimate of Ice-Sheet Mass Balance. Science 338, 1183–1189. doi:10.1126/science.1228102 Shepherd, A., Wingham, D., Payne, T., Skvarca, P., 2003. Larsen Ice Shelf Has Progressively Thinned. Science 302, 856–859. doi:10.1126/science.1089768 Shi, L., Allen, C.T., Ledford, J.R., Rodriguez-Morales, F., Blake, W.A., Panzer, B.G., Prokopiack, S.C., Leuschen, C.J., Gogineni, S., 2010. Multichannel Coherent Radar Depth Sounder for NASA Operation Ice Bridge. IEEE, pp. 1729–1732. doi:10.1109/IGARSS.2010.5649518 Shuman, C.A., Berthier, E., Scambos, T.A., 2011. 2001–2009 elevation and mass losses in the Larsen A and B embayments, Antarctic Peninsula. J. Glaciol. 57, 737–754. doi:10.3189/002214311797409811

104 References Skvarca, P., Rack, W., Rott, H., Donángelo, T.I., 1999. Climatic trend and the retreat and disintegration of ice shelves on the Antarctic Peninsula: an overview. Polar Res. 18, 151– 157. Skvarca, P., Rack, W., Rott, H., Ibarzábal y Donángelo, T., 1998. Evidence of recent climatic warming on the eastern Antarctic Peninsula. Ann. Glaciol. 27, 628–632. Skvarca, P., Rott, H., Nagler, T., 1995. Satellite imagery, a base line for glacier variation study on James Ross Island, Antarctica. Ann. Glaciol. 21, 291–296. Smith, R.C., Stammerjohn, S.E., 2001. Variations of surface air temperature and sea-ice extent in the western Antarctic Peninsula region. Ann. Glaciol. 33, 493–500. doi:10.3189/172756401781818662 Stammerjohn, S.E., Martinson, D.G., Smith, R.C., Yuan, X., Rind, D., 2008. Trends in Antarctic annual sea ice retreat and advance and their relation to El Niño–Southern Oscillation and Southern Annular Mode variability. J. Geophys. Res. Oceans 113, C03S90. doi:10.1029/2007JC004269 Stewart, J., 2011. Antarctica: an encyclopedia, 2nd ed. ed. McFarland & Co, Jefferson, N.C. Strozzi, T., Luckman, A., Murray, T., 2000. The evolution of a glacier surge observed with the ERS satellites, in: Proceedings of ERS-Envisat Symposium, ESA SP-461, CD-ROM, Gothenburg. Strozzi, T., Luckman, A., Murray, T., Wegmuller, U., Werner, C.L., 2002. Glacier motion estimation using SAR offset-tracking procedures. IEEE Trans. Geosci. Remote Sens. 40, 2384–2391. doi:10.1109/TGRS.2002.805079 Tapley, B.D., Bettadpur, S., Watkins, M., Reigber, C., 2004. The gravity recovery and climate experiment: Mission overview and early results: GRACE MISSION OVERVIEW AND EARLY RESULTS. Geophys. Res. Lett. 31, n/a-n/a. doi:10.1029/2004GL019920 Thomas, E.R., Marshall, G.J., McConnell, J.R., 2008. A doubling in snow accumulation in the western Antarctic Peninsula since 1850. Geophys. Res. Lett. 35. doi:10.1029/2007GL032529 Toutin, T., 2002. Three-dimensional topographic mapping with ASTER stereo data in rugged topography. IEEE Trans. Geosci. Remote Sens. 40, 2241–2247. doi:10.1109/TGRS.2002.802878 Turner, J., 2002. Spatial variability of Antarctic Peninsula net surface mass balance. J. Geophys. Res. 107. doi:10.1029/2001JD000755 Turner, J., Colwell, S.R., Marshall, G.J., Lachlan-Cope, T.A., Carleton, A.M., Jones, P.D., Lagun, V., Reid, P.A., Iagovkina, S., 2005. Antarctic climate change during the last 50 years. Int. J. Climatol. 25, 279–294. doi:10.1002/joc.1130 Turner, J., Lu, H., White, I., King, J.C., Phillips, T., Hosking, J.S., Bracegirdle, T.J., Marshall, G.J., Mulvaney, R., Deb, P., 2016. Absence of 21st century warming on Antarctic Peninsula consistent with natural variability. Nature 535, 411–415. doi:10.1038/nature18645 UNESCO, 1970. Combined heat, ice and water balances at selected glacier basins: a guide for compilation and assemblage of data for glaciermass balance measurements. Unesco, Lash, (Paris). van Lipzig, N.P.M., King, J.C., Lachlan-Cope, T.A., van den Broeke, M.R., 2004. Precipitation, sublimation, and snow drift in the Antarctic Peninsula region from a regional atmospheric model. J. Geophys. Res. Atmospheres 109, D24106. doi:10.1029/2004JD004701 van Lipzig, N.P.M., Meijgaard, E. van, Oerlemans, J., 2002. The spatial and temporal variability of the surface mass balance in Antarctica: results from a regional atmospheric climate model. Int. J. Climatol. 22, 1197–1217. doi:10.1002/joc.798 van Wessem, J.M., Reijmer, C.H., Morlighem, M., Mouginot, J., Rignot, E., Medley, B., Joughin, I., Wouters, B., Depoorter, M.A., Bamber, J.L., Lenaerts, J.T.M., De Van Berg, W.J., Van Den

105 References Broeke, M.R., Van Meijgaard, E., 2014. Improved representation of East Antarctic surface mass balance in a regional atmospheric climate model. J. Glaciol. 60, 761–770. doi:10.3189/2014JoG14J051 van Wessem, J.M., Ligtenberg, S.R.M., Reijmer, C.H., van de Berg, W.J., van den Broeke, M.R., Barrand, N.E., Thomas, E.R., Turner, J., Wuite, J., Scambos, T.A., van Meijgaard, E., 2016. The modelled surface mass balance of the Antarctic Peninsula at 5.5 km horizontal resolution. The Cryosphere 10, 271–285. doi:10.5194/tc-10-271-2016 Vaughan, D.G., Doake, C.S.M., 1996. Recent atmospheric warming and retreat of ice shelves on the Antarctic Peninsula. Nature 379, 328–331. doi:10.1038/379328a0 Vaughan, D.G., Marshall, G.J., Connolley, W.M., Parkinson, C., Mulvaney, R., Hodgson, D.A., King, J.C., Pudsey, C.J., Turner, J., 2003. Recent rapid regional climate warming on the Antarctic Peninsula. Clim. Change 60, 243–274. Veen, C.J. van der, 2002. Calving glaciers. Prog. Phys. Geogr. 26, 96–122. doi:10.1191/0309133302pp327ra Watson, R.C., 2009. Radar origins worldwide: history of its evolution in 13 nations through World War II. Trafford Publ, S.l. Wendt, J., Rivera, A., Wendt, A., Bown, F., Zamora, R., Casassa, G., Bravo, C., 2010. Recent ice- surface-elevation changes of Fleming Glacier in response to the removal of the Wordie Ice Shelf, Antarctic Peninsula. Ann. Glaciol. 51, 97–102. Wingham, D.J., Francis, C.R., Baker, S., Bouzinac, C., Brockley, D., Cullen, R., de Chateau- Thierry, P., Laxon, S.W., Mallow, U., Mavrocordatos, C., Phalippou, L., Ratier, G., Rey, L., Rostan, F., Viau, P., Wallis, D.W., 2006. CryoSat: A mission to determine the fluctuations in Earth’s land and marine ice fields. Adv. Space Res., Natural Hazards and Oceanographic Processes from Satellite Data 37, 841–871. doi:10.1016/j.asr.2005.07.027 Woodhouse, I.H., 2006. Introduction to microwave remote sensing. Taylor&Francis, Boca Raton. Wouters, B., Martin-Español, A., Helm, V., Flament, T., Wessem, J.M. van, Ligtenberg, S.R.M., Broeke, M.R. van den, Bamber, J.L., 2015. Dynamic thinning of glaciers on the Southern Antarctic Peninsula. Science 348, 899–903. doi:10.1126/science.aaa5727 Wuite, J., Rott, H., Hetzenecker, M., Floricioiu, D., De Rydt, J., Gudmundsson, G.H., Nagler, T., Kern, M., 2015. Evolution of surface velocities and ice discharge of Larsen B outlet glaciers from 1995 to 2013. The Cryosphere 9, 957–969. doi:10.5194/tc-9-957-2015 Zebker, H.A., Goldstein, R.M., 1986. Topographic mapping from interferometric synthetic aperture radar observations. J. Geophys. Res. Solid Earth 91, 4993–4999. doi:10.1029/JB091iB05p04993 Zwally, H.J., Giovinetto, M., Beckley, M.A., Saba, J.L., 2012a. Antarctic and Greenland Drainage Systems. Zwally, H.J., Giovinetto, M.B., 2011. Overview and Assessment of Antarctic Ice-Sheet Mass Balance Estimates: 1992–2009. Surv. Geophys. 32, 351–376. doi:10.1007/s10712-011- 9123-5 Zwally, H.J., Schutz, R., Bentley, C., Bufton, J., Herring, T., Minster, J., Spinhirne, J., Thomas, R., 2012b. GLAS/ICESat L2 Antarctic and Greenland Ice Sheet Altimetry Data, Version 33. Boulder, Colorado USA: NASA DAAC at the National Snow and Ice Data Center. [WWW Document]. URL http://nsidc.org/data/docs/daac/glas_icesat_l1_l2_global_altimetry.gd.html

106 Appendix

13 Appendix

13.1 Supplemental material to Study 1

Changes in ice dynamics, elevation and mass discharge of Dinsmoor-Bombardier- Edgeworth glacier system, Antarctic Peninsula

Seehaus, Thorsten1; Marinsek, Sebastián2,3; Helm, Veit4, Skvarca, Pedro5; Braun, Matthias1

1 Institut für Geographie, Universität Erlangen-Nürnberg, Wetterkreuz 15, D-91058 Erlangen, Germany 2 Instituto Antártico Argentino, Balcarce 290, C1064AAF, Buenos Aires, Argentina 3 Universidad Tecnológica Nacional, Facultad Regional Buenos Aires, Medrano 951, C1179AAQ, Buenos Aires, Argentina 4 Alfred-Wegener-Institut für Polar- und Meeresforschung, Am Alten Hafen 26, D-27568 Bremerhaven, Germany 5 Glaciarium, Museo del Hielo Patagónico, El Calafate 9405, Prov. Santa Cruz, Argentina

Overview:

S1 Surface mass balance approximations

S2 Post-processing and error estimation of intensity tracking results

S3 Processing of digital elevation models from TanDEM-X data

S4 Referencing and accuracy assessment of digital elevation models

S5 Hypsometric curve of measured surface elevation change

S6 Surface elevation profiles on Dinsmoor and Edgeworth Glacier

S7 Computation of ice discharge

S8 Mass balance 1995–2003 and 2003–2014

S9 Hypsometric curve of glacier catchment

107 Appendix

S1 Surface mass balance approximations

Five surface mass balance (bsfc) values were considered for the upper and lower bound approximation of bsfc at the catchment of Dinsmoor-Bombardier-Edgeworth Glacier (DBE):

 bsfc1 = 1750 kg/(m²a): Reported by Scambos et al. (2014) for DBE. It is the mean value of the RACMO-2.1/ANT (Lenaerts et al., 2012) regional climate model mass input for the time interval 1979–2011 and an ice covered catchment area of 822.7 km².

 bsfc2 = 1087±122 kg/(m²a): Estimated by Rott et al. (2011), at Crane Glacier (~160 km southwest of DBE) for the late 1990s.

 bsfc3 = 1047 kg/(m²a): Derived from the method described by Rott et al. (2011) for

calculating net accumulation in the nearby Larsen-B region; applied to DBE. Constant bsfc values were assumed below 200 m a.s.l. and above 1600 m a.s.l. with 400 kg/(m²a) and

1375 kg/(m²a) respectively. Between those elevations, a linear interpolation of bsfc was applied.

 bsfc4 = 1693 kg/(m²a): Calculated with the same method as for bsfc3, but with a bsfc of 2350 kg/(m²a) at regions above 1600 m a.s.l.—derived from a firn core analysis by Fernandoy et al. (2011) in 2010 at Plateau Laclavere (~140 km northeast of DBE, 1030 m a.s.l.).

 bsfc5 = 1530 kg/(m²a): It was assumed that the DBE was in equilibrium before 1995.

Following, bsfc6 was derived from the mean flux in 1992 and 1993 at the flux gate divided by the drainage area of 503 km².

The comparison of the bsfc values (1–5) led to the upper and lower bound approximations of bsfc(min) = 1070 kg/(m²a) (bsfc2, bsfc3) and bsfc(max) = 1720 kg/(m²a) (bsfc1, bsfc4).

S2 Post-processing and error estimation of intensity tracking results

The intensity tracking on single look complex SAR images was done in slant range geometry. The resulting offset fields were filtered for unreasonable values by using an efficient algorithm that compares the orientation and magnitude of a displacement vector to its surrounding vectors. It excludes >90% of erroneous displacement vectors (Burgess et al., 2012). The filtered displacement fields were transferred from slant into ground range displacement fields, considering topographic effects on the local incidence angle (Ф): d d = sr d ∗f gr sin ( ϕ ) sr gr (9)

with dgr being the displacement in ground range direction, dsr the displacement in slant range direction, Ф the local incident angle derived by means of the ASTER Global Digital Elevation Model of the Antarctic Peninsula (AP-DEM) (Cook et al., 2012), and satellite parameters. Finally the displacement fields were geocoded and orthorectified using the AP-DEM.

The error of the slant to ground range factor (σfgr) is given by the equation: 2 −cos ( ϕ ) 2 σ f = σ ϕ gr √(sin ( ϕ )2 ) (10)

108 Appendix

Figure 13-1 shows σfgr dependence on σФ (error of local incidence angle) and Ф. The surface slope on the glacier is quite gentle, so we assumed an error σФ = 0.5° over 300 m for the AP-DEM (Scambos et al., 2014). On the glacier surface, Ф is mostly within the interval 30° < Ф < 60°.

Consequently, σfgr ranges between 1.4% and 0.5% respectively, and was neglected. The influence of surface lowering on the slant range displacement was also minimal. It corresponded to 0.8 % of the 2014 median ice velocity at the terminus (1995–2014, mean dh/dt = 1.7 cm/d, Section 6.1.6.3).

Figure 13-1: Relative error of slant-to-ground range factor fgr. Dotted white lines indicate the assumed interval of local incidence angle Φ and its accuracy of σΦ = 0.5°.

c The uncertainties of the displacement fields caused by the coregistration process (σv ) were derived from 15 to 40 stable points on rock surfaces, which were not influenced by image distortions like SAR layover and shadowing (orange points, Fig. 13-3). The median of the displacement magnitude at these stable points was assigned to each offset field as the accuracy of the image coregistration of each analyzed SAR image pair.

T The uncertainties in the velocity fields due to the tracking algorithm (σv ) were estimated using the following formula, modified from McNabb et al. (2012): CΔx σT = v zΔt (11) where C is the uncertainty of the tracking algorithm [p], Δx the image resolution [m/p] , z the oversampling factor, and Δt the time interval [d] between the SAR images used. To improve the offset estimation accuracy, the patches were oversampled by a factor of two. We assumed an accuracy of the tracking algorithm of C = 0.4 p. The tracking was done in slant-range geometry without multilooking. The ground resolution of the input images was coarser in range direction. The range resolution in ground range geometry of the SAR images was taken as Δx (upper bound

109 Appendix approximation). For each sensor, a mean value of Δx, depending on the incidence angle and slant range resolution, was calculated at the study site and is listed below:  AMI SAR (ERS I/II): 16 m  R1 (Radarsat 1): 15 m  ASAR (ENVISAT): 16 m  PALSAR (ALOS): 7 m  TSX/TDX (TerraSAR-X/TanDEM-X): 2.5 m

C T We assumed that σv and σv are independent and estimated the total error of the obtained velocity C T fields σv as the sum of both. Table 13-1 lists the values Δt , σv , σv , and σv of each tracking result. T For the tracking results on March 30, 1996 and November 9, 1999, σv is quite large due to the very short time interval of one day between the SAR acquisitions. We excluded this value from our estimation of σv. Table 13-1 and Fig. 7-3a (in the manuscript) reveal that the error bars for the higher resolution X- band (TSX/TDX) and L-band (PALSAR) data are considerably smaller than for the coarser C-band systems of ERS-1/2, ENVISAT and RadarSAT-1.

110 Appendix

Table 13-1: Uncertainty σv of processed velocity fields. Date: Mean date of SAR acquisitions; Δt : C Time interval in days between repeat SAR acquisitions; σv : Uncertainty of image coregistration; T C σv : Uncertainty of intensity tracking process; 1996-03-30 and 1999-11-09: σv = σv see text above. The table is continued on the next page.

111 Appendix

112 Appendix S3 Processing of digital elevation models from TanDEM-X data

We generated digital elevation models (DEM) interferometrically from bistatic TanDEM-X (TDX) acquisitions. The SAR data was provided in Coregistered Single look Slant range Complex (CoSSC, DLR-IMF, 2012)) format by the German Space Agency (DLR). To simplify the phase unwrapping process of the interferogram and to reduce the potential of phase unwrapping errors, differential interferometric processing was applied. In a first step, the segment of the AP-DEM, which was covered by the TDX data, was transferred into the same coordinate system as the TDX data (slant range geometry). In this process, a cross- correlation was performed between the simulated backscatter intensity from the DEM in slant range geometry and the TDX backscatter intensity. Subsequently, an interferogram was simulated based on this DEM segment and the satellite orbit parameter of the TDX data. In the next step, a differential interferogram was formed from the two already coregistered SAR images of each TDX acquisition and the simulated interferogram. The differential interferogram was filtered and phase unwrapped, using a minimum cost flow algorithm within GAMMA Remote Sensing software. The unwrapped differential interferogram was then converted into a differential DEM using the phase- to-height sensitivity, derived from a simulated differential interferogram based on the orbit parameter of the TDX data, the AP-DEM, and the AP-DEM raised by 100 m. The AP-DEM elevations were subsequently re-added to the differential DEM to get an absolute TDX DEM related to the same reference surface as the AP-DEM (EGM96 geoid). The resulting DEM was first geocoded on the AP-DEM and finally orthorectified by its own elevation information.

S4 Referencing and accuracy assessment of digital elevation models

The stereoscopic DEMs from the level-2 standard NASA processing of Terra ASTER and SPOT SPIRIT project (Korona et al., 2009) were referenced to data from NASA's ICESat Geoscience Laser Altimeter System GLAS instrument (GLA06 L1B Global Elevation Data Version 33, Zwally et al., 2012). Only GLAS data, which was acquired within a time interval of ±20 days relative to the DEM acquisition, was considered for the accuracy assessment of the individual DEMs. The spatial coverage of the DEMs by the ICESat data is plotted in Fig. 13-2. The elevation offsets δh between the DEMs and the ICESat GLAS footprints were filtered for outliers of more than ±50 m. Each DEM was vertically adjusted by the mean of δh. We calculated the average elevation differences on nunataks (blue polygons in Fig. 13-3) to estimate the error of elevation differences (σh) relative to the SPOT DEM 2006. The values of σh and the number of ICESat footprints used for each DEM are listed in Table 13-2. CryoSat-2 SIRAL-2 (Helm et al., 2014) and OIB ATM laser altimeter data were used for absolute referencing and error assessment of the TDX DEMs. One flight line of OIB crossed the study site on 2011-11-14. A TDX data set was acquired on the same day and the processed TDX DEM was compared to the high quality laser scanning elevation measurements from ATM. The low SAR backscatter intensities at the OIB profile on that date indicated either a melting surface or bare ice. Thus the SAR penetration was considered to be quite low. The TDX DEM from 2011-11-14 was vertically adjusted by the median offset to the ATM data. On the ice mélange in front of the glacier, which is present on all TDX acquisitions, a median altitude of 1.24 m was measured on 2011-11- 14. All other TDX DEMs were referenced to that value on the ice mélange, including a correction

113 Appendix for the tidal signal. The amplitude of the tides for each data set was taken from the CATS tidal model CATS2008a_opt (Padman et al., 2002). Sufficient synchronous coverage (>7 spatial samples) of the TDX DEMs and CryoSat-2 measurements (±20 days) was only available for six dates over smooth topography (purple polygons and red points in Fig. 13-3). The offset measurements (σCS) between the TDX DEMs and CryoSat-2 data were obtained by using bilinear interpolation and filtered for outliers (>±50 m). The mean elevation difference of σCS is 0.82 m for all TDX DEMs. The authors are aware that the TDX X-band data has a larger penetration depth in dry snow and firn of up to several meters (Rignot et al., 2001) than in wet snow or glacier ice. This is certainly the case in higher elevation and regions with high accumulation, in particular reaching up to the AP plateau. However, we have no means to quantify the absolute errors in those areas.

Consequently, σCS was only measured on the lower parts of DBE. Since highest surface lowering rates are expected at areas below 1000 m a.s.l. (Scambos et al., 2014), the influence of DEM errors in higher elevations are most likely less important.

Table 13-2: Elevation data sets used to analyze surface elevation changes. σh : Mean vertical elevation difference on nunataks for ASTER and SPOT DEMs and mean vertical offset to CyoSat 2 measurements for TanDEM-X DEMs; n: number of altimeter footprints used for DEM referencing

From the observed σh and σCS values, we estimated a maximum error for our calculated elevation differences, relative to the ASTER DEM in 2003, of 5 m. Shuman et al. (2011) obtained the same value for elevation differences derived from ASTER and SPOT DEM at the Larsen-B embayment. This value is an upper bound approximation. DEM errors increase with steeper slopes (Toutin,

2002), but the offset σh between the stereoscopic DEMs could only be estimated on nunataks, where the surface slopes are typically higher than on the glacier. Furthermore, the offsets δCS between CryoSat-2 measurements and TDX DEMs also include the inaccuracy of the CryoSat-2 measurements. Note: All of the elevation dataset discussed above were referenced to the EGM96 geoid model.

114 Appendix

Figure 13-2: Spatial coverage of DEM 95 (subset area S3) and of TanDEM-X DEMs by CryoSat 2 data (masked to areas with smooth topography), areas on nunataks used for accuracy assessment of stereoscopic DEMs and stable points used to estimate the coregistration error of intensity tracking results.

Figure 13-3: Spatial extents of the stereoscopic DEMs and their coverage by ICESat GLAS data (same color as the DEM extents of the respective dates) acquired in a time interval of ±20 days relative to the DEM acquisition (date format: yyyy-mm-dd).

115 Appendix S5 Hypsometric curves of measured surface elevation change

Figure 13-4: Hypsometric curves of the surface elevation change on DBE below 1000 m a.s.l. 2003–2014: Surface elevation changes were obtained on subset area S1 from the ASTER DEM in 2003 and the TDX DEM on 21st. March 2014; 1995–2003: Surface elevation changes at altitudes above 550 m (not covered by the DEM95—subset area S3, Figure 13-2) were derived by linear interpolation towards 0 m at 1000 m a.s.l., considering the very low surface lowering rate of 0.23 m/a found by Scambos et al. (2014) for DBE at areas above 1000 m a.s.l.; 1995–2014: Surface elevation changes are the sum of the changes in the periods 1995–2003 and 2003–2014. Dashed lines represent the mean surface elevation change at areas below 1000 m a.s.l. of −100.8 m (1995–2003, −11.5 m/y), −17.3 m (2003–2014, −1.7 m/y) and −118.1 m (1995–2014, −6.2 m/y)

116 Appendix S6 Surface elevation profiles on Dinsmoor and Edgeworth Glacier

Figure 13-5: Ice surface elevation profiles on Dinsmoor and Edgeworth Glacier, from terminus to higher elevations (see Figure 7-4 in the main manuscript). Date format is yyyy-mm-dd.

117 Appendix

S7 Computation of ice discharge

The ice flux was calculated at the Operation IceBridge (OIB) profile (orange line in Figure 7-1) covering all three glaciers. Bedrock elevation from the three available sources (Bedmap2, Fretwell et al., 2013; OIB Multichannel Coherent Radar Depth Sounder (MCoRDS) on November 14, 2011, Leuschen et al., 2010; Huss and Farinotti, 2014) was used to estimate the ice-thickness on the Dinsmoor-Bombardier-Edgeworth Glacier system (DBE). Figure 13-1 shows the different bedrock elevations along the OIB Airborne Topographic Mapper (ATM) (Krabill, 2010) profile. The MCoRDS measurements only revealed bedrock reflections in the central part of Edgeworth Glacier. For comparison with the other data sets, the MCoRDS profile was linearly extrapolated towards the glacier margins assuming constant slope on both sides of the valley. Particularly the Bedmap2 bedrock elevation is higher than the ATM surface elevation over large parts of Dinsmoor and Bombardier Glacier. As a consequence, the data set from Huss and Farinotti (2014) is considered to be the best bedrock elevation map currently available for the region.

The ice flux was approximated as the sum over the ice flux fj at smaller cross sections j along the profile:

F=∑ 0 .96∗ρi∗v j∗t j∗w j∗sin ( α j)=∑ f j j j (12)

where ρi is the assumed mean ice density of 900 kg/m³, vj the surface velocity magnitude [m/d] from the SAR tracking results, αj the angle [°] between the flow direction and the cross section j, tj the ice-thickness [m], and wj the width [m] of the column. The mean flow velocity of the column was approximated to be 0.96 vj (Rott et al., 2011) for grounded ice. The flow direction values were additionally filtered by removing αj that differed more than 20° from a manually defined flow direction profile along the flux gate. Gaps in vj were linearly interpolated. Variations in the ice- thickness at the flux gate caused by surface elevation changes were taken into account. The interpolated surface lowering curve (Section 6.1.6.3) was used to calculate the temporal changes in tj. Input-value errors are supposed to be independent. Consequently, the mass-flux error σF results from 2 2 2 ∂ f j 2 ∂ f j 2 ∂ f j 2 σ F=∑ σ t + σv + σα ∂t j ∂v j ∂ α j (13) j √( j ) ( j ) ( j )

The σvj was estimated as in Section S2 and σαj was assumed to be 0.35 rad (20°). The error of the ice-thickness (σtj) was estimated to be 10 m and 30 m for ice-thickness based on Huss and Farinotti (2014) and the Bedmap2 bedrock data set, respectively. These values were assessed using each data set’s uncertainty map. Inaccuracies of the ice density and column width with were not considered.

118 Appendix

Figure 13-6: Ice surface and bedrock elevation data along the Operation IceBridge (OIB) ATM Nadir profile on DBE. Red line: Bedrock elevation from Huss and Farinotti (2014). Blue line: Bedrock elevation from Bedmap2. Black line: Surface elevation from OIB ATM laser scanner, Track ID = 0 (Nadir) Green line: Bedrock elevation calculated from OIB ATM surface elevation data and MCoRDS L2 ice thickness data on DBE. Green dotted line: Linear interpolation of OIB MCoRDS bedrock profile toward the margins of Edgeworth Glacier.

119 Appendix

S8 Mass balance 1995–2003 and 2003–2014

Table 13-3: a) Total ice mass balance ΔM and b) contribution to sea level rise ΔMslr between 1995 and 2003 for different boundary conditions and methods. See Section 6.1.5.5 in the manuscript for detailed description of input parameters. Uncertainties are given with one σ.

Table 13-4: a) Total ice mass balance ΔM and b) contribution to sea level rise ΔMslr between 2003 and 2014 for different boundary conditions and methods. See Section 6.1.5.5 in the main manuscript for detailed description of input parameters. Uncertainties are given with one σ.

120 Appendix

S9 Hypsometric curve of glacier catchment

Figure 13-7: Hypsometric curve of Dinsmoor-Bombardier-Edgeworth Glacier catchment (see Fig. 7-1 in the manuscript) based on the digital elevation model from Cook et al. 2012.

121 Appendix

13.2 Supplemental material to Study 3

Detailed analysis of changes in glacier dynamics on the northern Antarctic Peninsula since 1985

Seehaus Thorsten1*, Cook Alison2, Barbosa Aline3, van Wessem Jan Melchior4, Reijmer Carleen H.4, Marinsek Sebastián5, Braun Matthias1

1 Institute of Geography, Friedrich-Alexander-University Erlangen-Nuremberg, Wetterkreuz 15, D- 91058 Erlangen, Germany 2 Department of Geography, Durham University, United Kingdom 3 Laboratório de Monitoramento da Criosfera, Fundação Universidade Federal do Rio Grande, Brazil 4 Institute for Marine and Atmospheric Research Utrecht, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands 5 Instituto Antártico Argentino, Balcarce 290, C1064AAF, Buenos Aires, Argentina

Overview:

Table 13-5: Uncertainties of intensity tracking results

Table 13-6: Observe parameters of the individual glaciers

Figure S1-S74: Changes in glacier area and flow speed

122 Appendix

Table 13-5: Uncertainty σv of intensity tracking results. Date: Mean date of SAR acquisitions; Δt : Time C T interval in days between consecutive SAR acquisitions; σv : Uncertainty of image coregistration; σv : C Uncertainty of intensity tracking process; If Δt=1d -> σv = σv see manuscript. Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 1992-12-25 ERS 35 0.13 9721 0.05 0.14 1992-12-25 ERS 35 0.25 23678 0.05 0.26 1993-01-12 ERS 70 0.07 9880 0.02 0.07 1993-01-29 ERS 35 0.10 6090 0.05 0.11 1993-01-29 ERS 35 0.23 4533 0.05 0.24 1993-02-01 ERS 35 0.20 6321 0.05 0.21 1994-02-01 ERS 21 0.35 22007 0.08 0.36 1994-02-18 ERS 54 0.07 28834 0.03 0.08 1994-02-28 ERS 33 0.16 26276 0.05 0.17 1995-10-31 ERS 1 0.41 150 1.60 0.41 1995-11-14 ERS 1 0.36 1961 1.60 0.36 1995-11-16 ERS 1 0.29 448 1.60 0.29 1995-12-18 ERS 71 0.02 68711 0.02 0.03 1995-12-18 ERS 70 0.03 77246 0.02 0.04 1995-12-19 ERS 71 0.02 70974 0.02 0.03 1995-12-19 ERS 70 0.06 67287 0.02 0.06 1995-12-19 ERS 69 0.12 66877 0.02 0.12 1995-12-20 ERS 70 0.04 70897 0.02 0.04 1995-12-21 ERS 70 0.08 10755 0.02 0.08 1995-12-21 ERS 69 0.09 9000 0.02 0.10 1996-01-22 ERS 1 0.24 49973 1.60 0.24 1996-01-23 ERS 1 0.34 546 1.60 0.34 1996-02-11 ERS 35 0.12 10215 0.05 0.12 1996-02-11 ERS 35 0.14 8164 0.05 0.15 1996-02-13 ERS 35 0.06 23882 0.05 0.08 1996-02-15 ERS 35 0.14 9379 0.05 0.15 1996-02-29 ERS 35 0.02 39573 0.05 0.05 1996-03-03 ERS 34 0.05 18324 0.05 0.07 1996-03-03 ERS 35 0.05 18395 0.05 0.07 1996-03-20 ERS 1 0.30 9049 1.60 0.30 1997-02-13 ERS 35 0.04 44246 0.05 0.06 1997-02-15 ERS 35 0.11 14969 0.05 0.12 1997-02-18 ERS 35 0.09 6705 0.05 0.10 1998-02-03 ERS 35 0.07 3176 0.05 0.08 1999-11-09 ERS 1 0.34 4022 1.60 0.34 2002-02-07 ERS 35 0.07 9893 0.05 0.09 2002-11-29 ERS 35 0.13 61073 0.05 0.13 2002-12-03 ERS 35 0.13 19079 0.05 0.13 2002-12-08 ERS 35 0.29 1965 0.05 0.29 2002-12-21 ERS 70 0.05 21331 0.02 0.05 2002-12-21 ERS 35 0.27 3396 0.05 0.27 2002-12-26 ERS 70 0.13 2437 0.02 0.13 2003-01-07 ERS 35 0.05 24658 0.05 0.07 2003-01-08 ERS 70 0.19 4794 0.02 0.19 2003-01-12 ERS 35 0.09 2548 0.05 0.10 2003-01-25 ERS 35 0.10 14207 0.05 0.11 2004-11-01 ERS 35 0.17 30346 0.05 0.17 2004-11-17 ERS 70 0.06 71277 0.02 0.07 2004-11-19 ERS 70 0.08 32153 0.02 0.09 2004-12-06 ERS 35 0.11 33520 0.05 0.12 2004-12-24 ERS 70 0.11 34409 0.02 0.11 2004-12-25 ERS 35 0.14 12592 0.05 0.14 2005-01-10 ERS 35 0.28 23466 0.05 0.28

123 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2006-11-03 ERS 35 0.19 56628 0.05 0.19 2006-11-04 ERS 35 0.14 70277 0.05 0.14 2008-10-29 ERS 35 0.07 9881 0.05 0.08 2010-02-08 ERS 35 0.18 18041 0.05 0.19 2010-02-26 ERS 70 0.11 19172 0.02 0.11 2010-03-15 ERS 35 0.10 23486 0.05 0.11 2000-09-22 R1 24 0.10 20810 0.06 0.12 2000-09-22 R1 24 0.14 33870 0.06 0.15 2000-10-01 R1 24 0.06 30397 0.06 0.09 2006-08-22 R1 24 0.07 57259 0.06 0.10 2006-08-22 R1 24 0.08 21635 0.06 0.10 2003-12-22 ENVISAT 35 0.31 38866 0.05 0.31 2004-01-09 ENVISAT 70 0.03 61495 0.02 0.04 2004-01-10 ENVISAT 35 0.13 1790 0.05 0.13 2004-01-28 ENVISAT 70 0.16 1510 0.02 0.16 2004-02-14 ENVISAT 35 0.09 1898 0.05 0.10 2004-03-20 ENVISAT 35 0.13 3299 0.05 0.14 2004-04-24 ENVISAT 35 0.12 3505 0.05 0.13 2004-05-29 ENVISAT 35 0.10 3623 0.05 0.11 2004-07-03 ENVISAT 35 0.10 3546 0.05 0.11 2004-07-19 ENVISAT 35 0.03 60612 0.05 0.06 2004-08-07 ENVISAT 35 0.11 3418 0.05 0.12 2004-09-11 ENVISAT 35 0.14 3400 0.05 0.15 2004-10-16 ENVISAT 35 0.15 3449 0.05 0.16 2004-12-06 ENVISAT 35 0.06 63965 0.05 0.08 2005-01-28 ENVISAT 70 0.02 62239 0.02 0.03 2005-03-05 ENVISAT 35 0.15 2744 0.05 0.15 2005-03-21 ENVISAT 35 0.19 64254 0.05 0.19 2005-04-09 ENVISAT 35 0.13 2904 0.05 0.14 2005-05-14 ENVISAT 35 0.17 3016 0.05 0.17 2005-06-18 ENVISAT 35 0.13 3631 0.05 0.14 2005-07-23 ENVISAT 35 0.14 2943 0.05 0.14 2005-08-08 ENVISAT 35 0.12 68061 0.05 0.13 2006-02-15 ENVISAT 35 0.07 61205 0.05 0.08 2006-03-25 ENVISAT 35 0.14 2755 0.05 0.15 2006-07-08 ENVISAT 35 0.08 3488 0.05 0.09 2006-08-09 ENVISAT 35 0.06 60954 0.05 0.08 2006-08-12 ENVISAT 35 0.15 3302 0.05 0.15 2006-09-16 ENVISAT 35 0.14 3295 0.05 0.15 2006-10-21 ENVISAT 35 0.16 2741 0.05 0.17 2007-02-18 ENVISAT 70 0.03 71538 0.02 0.04 2007-04-29 ENVISAT 70 0.04 65692 0.02 0.05 2007-06-20 ENVISAT 35 0.03 63862 0.05 0.05 2007-08-12 ENVISAT 70 0.04 61079 0.02 0.05 2007-09-01 ENVISAT 35 0.15 3391 0.05 0.16 2007-10-03 ENVISAT 35 0.10 61336 0.05 0.11 2007-10-06 ENVISAT 35 0.16 3255 0.05 0.16 2008-04-30 ENVISAT 35 0.10 63576 0.05 0.11 2008-06-22 ENVISAT 70 0.03 57922 0.02 0.04 2008-08-13 ENVISAT 35 0.07 60539 0.05 0.08 2009-03-11 ENVISAT 35 0.11 64638 0.05 0.12 2009-07-29 ENVISAT 35 0.03 61130 0.05 0.05 2006-06-10 ALOS 46 0.02 15503 0.02 0.02 2006-06-17 ALOS 46 0.01 61958 0.02 0.02 2006-06-25 ALOS 46 0.08 581 0.02 0.09 2006-07-14 ALOS 46 0.02 9476 0.02 0.02 2006-09-21 ALOS 92 0.02 9912 0.01 0.02

124 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2006-12-23 ALOS 46 0.08 5135 0.02 0.08 2007-12-04 ALOS 46 0.03 10220 0.02 0.04 2007-12-14 ALOS 46 0.04 2193 0.02 0.04 2008-05-14 ALOS 46 0.01 43889 0.02 0.02 2008-10-21 ALOS 46 0.02 10711 0.02 0.02 2008-10-31 ALOS 46 0.13 2461 0.02 0.13 2008-11-13 ALOS 92 0.02 10861 0.01 0.02 2008-11-14 ALOS 46 0.02 33136 0.02 0.02 2008-12-06 ALOS 46 0.04 10213 0.02 0.04 2008-12-07 ALOS 92 0.02 36230 0.01 0.02 2008-12-16 ALOS 46 0.07 2291 0.02 0.07 2008-12-29 ALOS 92 0.02 10998 0.01 0.02 2008-12-30 ALOS 46 0.04 37661 0.02 0.04 2009-01-21 ALOS 46 0.02 10677 0.02 0.03 2009-12-02 ALOS 46 0.05 3484 0.02 0.05 2009-12-09 ALOS 46 0.03 9707 0.02 0.03 2009-12-21 ALOS 46 0.05 2455 0.02 0.05 2009-12-26 ALOS 46 0.03 9385 0.02 0.03 2010-01-19 ALOS 46 0.02 15505 0.02 0.02 2010-10-08 ALOS 46 0.04 620 0.02 0.04 2010-10-17 ALOS 46 0.03 79294 0.02 0.03 2010-11-06 ALOS 46 0.08 2212 0.02 0.08 2010-11-08 ALOS 46 0.01 16076 0.02 0.02 2010-11-10 ALOS 46 0.02 422 0.02 0.03 2010-11-13 ALOS 46 0.04 9956 0.02 0.05 2010-11-29 ALOS 92 0.03 2069 0.01 0.03 2010-12-01 ALOS 92 0.01 18027 0.01 0.01 2010-12-03 ALOS 92 0.40 426 0.01 0.40 2010-12-06 ALOS 92 0.03 10352 0.01 0.03 2010-12-11 ALOS 92 0.04 4683 0.01 0.04 2010-12-12 ALOS 46 0.03 9480 0.02 0.04 2010-12-22 ALOS 46 0.05 1992 0.02 0.05 2010-12-26 ALOS 46 0.02 411 0.02 0.03 2010-12-29 ALOS 46 0.03 10478 0.02 0.04 2010-12-31 ALOS 46 0.01 46824 0.02 0.02 2011-01-18 ALOS 92 0.16 430 0.01 0.16 2011-02-08 ALOS 46 0.01 17569 0.02 0.02 2011-02-10 ALOS 46 0.01 394 0.02 0.02 2008-10-19 TSX/TDX 11 0.05 4560 0.02 0.05 2008-10-25 TSX/TDX 22 0.02 4362 0.01 0.02 2008-10-30 TSX/TDX 11 0.03 4507 0.02 0.04 2009-08-01 TSX/TDX 11 0.02 11170 0.02 0.03 2009-10-28 TSX/TDX 11 0.06 4220 0.02 0.07 2010-10-26 TSX/TDX 33 0.02 2678 0.01 0.02 2010-11-01 TSX/TDX 44 0.02 3442 0.01 0.02 2010-11-17 TSX/TDX 22 0.01 5995 0.01 0.01 2010-11-17 TSX/TDX 11 0.06 3599 0.02 0.07 2010-11-28 TSX/TDX 99 0.01 3063 0.00 0.01 2010-12-15 TSX/TDX 66 0.02 3476 0.00 0.02 2010-12-20 TSX/TDX 77 0.01 3524 0.00 0.01 2010-12-20 TSX/TDX 55 0.01 4297 0.00 0.02 2010-12-26 TSX/TDX 66 0.01 4341 0.00 0.01 2011-01-22 TSX/TDX 11 0.02 4722 0.02 0.03 2011-06-25 TSX/TDX 22 0.01 15556 0.01 0.02 2011-06-25 TSX/TDX 22 0.04 9886 0.01 0.04 2011-07-06 TSX/TDX 44 0.04 10380 0.01 0.04 2011-07-16 TSX/TDX 22 0.04 3582 0.01 0.04

125 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2011-07-17 TSX/TDX 22 0.01 15712 0.01 0.02 2011-07-16 TSX/TDX 22 0.10 1421 0.01 0.10 2011-07-17 TSX/TDX 22 0.03 10450 0.01 0.03 2011-07-28 TSX/TDX 44 0.02 10607 0.01 0.02 2011-08-03 TSX/TDX 22 0.40 614 0.01 0.40 2011-08-08 TSX/TDX 22 0.03 10394 0.01 0.04 2011-08-14 TSX/TDX 44 0.14 1556 0.01 0.14 2011-08-19 TSX/TDX 44 0.03 10054 0.01 0.03 2011-08-19 TSX/TDX 55 0.04 2385 0.00 0.04 2011-08-24 TSX/TDX 22 0.03 1894 0.01 0.03 2011-08-24 TSX/TDX 55 0.03 10578 0.00 0.03 2011-08-29 TSX/TDX 33 0.03 1856 0.01 0.03 2011-08-30 TSX/TDX 22 0.02 15605 0.01 0.02 2011-08-30 TSX/TDX 22 0.06 7157 0.01 0.06 2011-09-04 TSX/TDX 33 0.01 15878 0.01 0.01 2011-09-09 TSX/TDX 11 0.06 2325 0.02 0.06 2011-09-14 TSX/TDX 11 0.05 3667 0.02 0.05 2011-09-14 TSX/TDX 11 0.12 1279 0.02 0.12 2011-09-15 TSX/TDX 11 0.03 15546 0.02 0.03 2011-09-15 TSX/TDX 11 0.07 7819 0.02 0.07 2011-09-27 TSX/TDX 44 0.14 2001 0.01 0.14 2011-10-01 TSX/TDX 33 0.02 1956 0.01 0.02 2011-10-01 TSX/TDX 44 0.04 3582 0.01 0.04 2011-10-06 TSX/TDX 33 0.04 3602 0.01 0.05 2011-10-06 TSX/TDX 33 0.11 1353 0.01 0.11 2011-10-12 TSX/TDX 66 0.02 3453 0.00 0.02 2011-10-17 TSX/TDX 55 0.03 3541 0.00 0.03 2011-10-23 TSX/TDX 11 0.06 2018 0.02 0.06 2011-11-03 TSX/TDX 22 0.05 3533 0.01 0.05 2011-11-03 TSX/TDX 22 0.07 1209 0.01 0.07 2011-11-25 TSX/TDX 22 0.03 3507 0.01 0.03 2011-12-06 TSX/TDX 11 0.06 2432 0.02 0.06 2011-12-12 TSX/TDX 33 0.01 13467 0.01 0.01 2011-12-13 TSX/TDX 44 0.05 2328 0.01 0.05 2011-12-17 TSX/TDX 22 0.01 4172 0.01 0.02 2011-12-18 TSX/TDX 33 0.08 2365 0.01 0.08 2012-01-03 TSX/TDX 11 0.01 16220 0.02 0.03 2012-01-03 TSX/TDX 11 0.07 8576 0.02 0.07 2012-01-31 TSX/TDX 55 0.05 2338 0.00 0.05 2012-03-09 TSX/TDX 11 0.02 13279 0.02 0.03 2012-03-09 TSX/TDX 11 0.16 7483 0.02 0.16 2012-03-10 TSX/TDX 22 0.07 2343 0.01 0.07 2012-03-15 TSX/TDX 22 0.01 15451 0.01 0.01 2012-03-15 TSX/TDX 33 0.05 2290 0.01 0.05 2012-03-15 TSX/TDX 22 0.07 7142 0.01 0.07 2012-03-20 TSX/TDX 11 0.08 6422 0.02 0.08 2012-03-21 TSX/TDX 44 0.05 2265 0.01 0.05 2012-03-25 TSX/TDX 22 0.11 1258 0.01 0.11 2012-03-26 TSX/TDX 55 0.05 2143 0.00 0.05 2012-03-26 TSX/TDX 11 0.19 2259 0.02 0.19 2012-04-01 TSX/TDX 22 0.14 2362 0.01 0.14 2012-04-06 TSX/TDX 33 0.06 2248 0.01 0.06 2012-04-06 TSX/TDX 11 0.10 2316 0.02 0.10 2012-04-12 TSX/TDX 22 0.05 2100 0.01 0.05 2012-04-17 TSX/TDX 22 0.02 15486 0.01 0.02 2012-04-17 TSX/TDX 22 0.05 7244 0.01 0.05 2012-04-30 TSX/TDX 11 0.04 1747 0.02 0.05

126 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2012-05-08 TSX/TDX 66 0.02 3381 0.00 0.02 2012-05-09 TSX/TDX 22 0.02 15305 0.01 0.02 2012-05-09 TSX/TDX 55 0.04 2344 0.00 0.04 2012-05-09 TSX/TDX 22 0.05 6241 0.01 0.05 2012-05-13 TSX/TDX 77 0.02 3656 0.00 0.02 2012-05-15 TSX/TDX 44 0.04 2221 0.01 0.04 2012-05-19 TSX/TDX 22 0.03 3672 0.01 0.03 2012-05-19 TSX/TDX 22 0.10 1275 0.01 0.10 2012-05-20 TSX/TDX 55 0.04 2375 0.00 0.04 2012-05-24 TSX/TDX 33 0.04 1210 0.01 0.04 2012-05-30 TSX/TDX 33 0.03 2544 0.01 0.03 2012-06-04 TSX/TDX 11 0.05 3532 0.02 0.06 2012-06-04 TSX/TDX 11 0.10 1351 0.02 0.11 2012-06-05 TSX/TDX 33 0.01 15558 0.01 0.01 2012-06-11 TSX/TDX 11 0.09 2222 0.02 0.09 2012-06-15 TSX/TDX 11 0.08 3328 0.02 0.09 2012-06-15 TSX/TDX 11 0.10 1280 0.02 0.10 2012-06-21 TSX/TDX 11 0.07 2621 0.02 0.07 2012-06-27 TSX/TDX 11 0.06 7647 0.02 0.06 2012-06-28 TSX/TDX 44 0.04 2293 0.01 0.04 2012-07-03 TSX/TDX 55 0.04 2350 0.00 0.04 2012-07-03 TSX/TDX 33 0.05 2292 0.01 0.05 2012-07-09 TSX/TDX 44 0.04 2389 0.01 0.04 2012-07-13 TSX/TDX 33 0.03 2765 0.01 0.03 2012-07-19 TSX/TDX 33 0.02 15662 0.01 0.02 2012-07-25 TSX/TDX 11 0.09 2122 0.02 0.09 2012-08-04 TSX/TDX 11 0.07 2545 0.02 0.07 2012-08-09 TSX/TDX 11 0.07 3577 0.02 0.07 2012-08-09 TSX/TDX 11 0.12 1204 0.02 0.13 2012-08-10 TSX/TDX 11 0.07 7151 0.02 0.07 2012-08-11 TSX/TDX 44 0.08 2444 0.01 0.08 2012-08-16 TSX/TDX 55 0.04 2374 0.00 0.04 2012-08-22 TSX/TDX 44 0.04 2230 0.01 0.04 2012-09-07 TSX/TDX 11 0.14 1690 0.02 0.14 2012-09-23 TSX/TDX 33 0.05 1078 0.01 0.05 2012-09-29 TSX/TDX 55 0.04 1597 0.00 0.04 2012-09-29 TSX/TDX 33 0.06 2397 0.01 0.06 2012-10-05 TSX/TDX 44 0.08 2401 0.01 0.08 2012-10-10 TSX/TDX 55 0.05 2372 0.00 0.05 2012-10-20 TSX/TDX 33 0.03 2520 0.01 0.03 2012-10-21 TSX/TDX 11 0.09 2179 0.02 0.09 2012-10-27 TSX/TDX 22 0.08 2296 0.01 0.08 2012-11-01 TSX/TDX 11 0.10 2327 0.02 0.10 2012-11-01 TSX/TDX 33 0.17 1923 0.01 0.17 2012-11-05 TSX/TDX 11 0.05 3446 0.02 0.05 2012-11-05 TSX/TDX 11 0.13 1186 0.02 0.13 2012-11-07 TSX/TDX 44 0.05 2312 0.01 0.05 2012-11-12 TSX/TDX 33 0.05 2364 0.01 0.06 2012-11-12 TSX/TDX 11 0.12 2354 0.02 0.12 2012-11-18 TSX/TDX 22 0.07 2419 0.01 0.07 2012-11-23 TSX/TDX 11 0.08 2204 0.02 0.09 2012-12-26 TSX/TDX 55 0.03 2141 0.00 0.03 2013-02-23 TSX/TDX 77 0.01 3503 0.00 0.01 2013-03-01 TSX/TDX 11 0.08 2802 0.02 0.08 2013-03-17 TSX/TDX 11 0.06 3749 0.02 0.07 2013-03-17 TSX/TDX 11 0.14 1255 0.02 0.14 2013-03-23 TSX/TDX 22 0.03 3632 0.01 0.03

127 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2013-03-23 TSX/TDX 22 0.08 1196 0.01 0.08 2013-03-26 TSX/TDX 11 0.08 1992 0.02 0.08 2013-03-28 TSX/TDX 11 0.17 1347 0.02 0.18 2013-03-29 TSX/TDX 33 0.05 1148 0.01 0.05 2013-04-03 TSX/TDX 33 0.09 2117 0.01 0.09 2013-04-10 TSX/TDX 22 0.06 2172 0.01 0.07 2013-04-15 TSX/TDX 33 0.07 2237 0.01 0.07 2013-04-26 TSX/TDX 55 0.05 2275 0.00 0.05 2013-04-26 TSX/TDX 11 0.12 2379 0.02 0.13 2013-04-30 TSX/TDX 55 0.02 3261 0.00 0.03 2013-06-08 TSX/TDX 22 0.03 3820 0.01 0.03 2013-06-08 TSX/TDX 22 0.04 1021 0.01 0.04 2013-06-19 TSX/TDX 44 0.02 3719 0.01 0.02 2013-06-30 TSX/TDX 22 0.03 3813 0.01 0.03 2013-06-30 TSX/TDX 22 0.09 1258 0.01 0.09 2013-07-28 TSX/TDX 33 0.01 15233 0.01 0.02 2013-08-02 TSX/TDX 33 0.02 2763 0.01 0.02 2013-08-25 TSX/TDX 33 0.05 2311 0.01 0.05 2013-08-30 TSX/TDX 33 0.01 15399 0.01 0.01 2013-09-20 TSX/TDX 33 0.03 3602 0.01 0.03 2013-09-20 TSX/TDX 33 0.05 1292 0.01 0.05 2013-09-27 TSX/TDX 33 0.04 2235 0.01 0.04 2013-10-02 TSX/TDX 33 0.01 15262 0.01 0.01 2013-10-23 TSX/TDX 33 0.02 3578 0.01 0.02 2013-10-23 TSX/TDX 33 0.05 1283 0.01 0.05 2013-10-30 TSX/TDX 33 0.05 2317 0.01 0.05 2013-11-02 TSX/TDX 11 0.02 9090 0.02 0.03 2013-11-02 TSX/TDX 11 0.07 484 0.02 0.07 2013-11-04 TSX/TDX 33 0.02 15102 0.01 0.02 2013-11-09 TSX/TDX 11 0.05 2652 0.02 0.06 2013-11-10 TSX/TDX 55 0.04 2294 0.00 0.04 2013-11-15 TSX/TDX 22 0.04 2878 0.01 0.05 2013-11-20 TSX/TDX 22 0.03 3538 0.01 0.04 2013-11-20 TSX/TDX 33 0.04 2955 0.01 0.04 2013-11-20 TSX/TDX 11 0.08 2846 0.02 0.08 2013-11-20 TSX/TDX 22 0.10 1321 0.01 0.10 2013-11-21 TSX/TDX 11 0.08 2180 0.02 0.08 2013-11-25 TSX/TDX 33 0.02 3312 0.01 0.02 2013-11-25 TSX/TDX 33 0.05 1125 0.01 0.05 2013-11-26 TSX/TDX 11 0.03 15060 0.02 0.03 2013-11-26 TSX/TDX 22 0.04 2825 0.01 0.04 2013-11-26 TSX/TDX 11 0.08 6708 0.02 0.09 2013-11-27 TSX/TDX 22 0.08 2346 0.01 0.09 2013-11-30 TSX/TDX 44 0.00 8207 0.01 0.01 2013-12-01 TSX/TDX 44 0.02 3438 0.01 0.02 2013-12-01 TSX/TDX 33 0.03 2670 0.01 0.03 2013-12-01 TSX/TDX 11 0.06 2893 0.02 0.06 2013-12-02 TSX/TDX 22 0.01 14680 0.01 0.01 2013-12-02 TSX/TDX 33 0.04 2079 0.01 0.04 2013-12-02 TSX/TDX 22 0.06 6620 0.01 0.06 2013-12-02 TSX/TDX 11 0.23 1957 0.02 0.24 2013-12-06 TSX/TDX 11 0.05 3548 0.02 0.06 2013-12-06 TSX/TDX 11 0.15 1322 0.02 0.15 2013-12-07 TSX/TDX 11 0.02 14924 0.02 0.03 2013-12-07 TSX/TDX 22 0.04 2905 0.01 0.04 2013-12-07 TSX/TDX 11 0.11 8347 0.02 0.11 2013-12-08 TSX/TDX 22 0.08 2021 0.01 0.08

128 Appendix Date Δt σ C σ T σ Satellite v n v v [yyyy-mm-dd] [d] [m/d] [m/d] [m/d] 2013-12-12 TSX/TDX 22 0.03 3508 0.01 0.03 2013-12-12 TSX/TDX 33 0.03 2814 0.01 0.03 2013-12-12 TSX/TDX 11 0.07 3039 0.02 0.08 2013-12-12 TSX/TDX 22 0.09 1242 0.01 0.09 2013-12-13 TSX/TDX 33 0.06 2306 0.01 0.06 2013-12-13 TSX/TDX 11 0.07 2024 0.02 0.08 2013-12-17 TSX/TDX 11 0.02 3978 0.02 0.03 2013-12-17 TSX/TDX 33 0.03 3323 0.01 0.03 2013-12-17 TSX/TDX 11 0.14 1290 0.02 0.14 2013-12-18 TSX/TDX 33 0.01 13920 0.01 0.01 2013-12-18 TSX/TDX 22 0.03 2741 0.01 0.04 2013-12-23 TSX/TDX 22 0.03 3725 0.01 0.03 2013-12-23 TSX/TDX 11 0.05 2877 0.02 0.06 2013-12-23 TSX/TDX 22 0.09 1118 0.01 0.10 2013-12-24 TSX/TDX 22 0.01 14893 0.01 0.01 2013-12-24 TSX/TDX 22 0.05 7587 0.01 0.05 2013-12-24 TSX/TDX 11 0.09 2342 0.02 0.09 2013-12-28 TSX/TDX 11 0.05 3475 0.02 0.05 2013-12-28 TSX/TDX 11 0.14 1096 0.02 0.15 2013-12-30 TSX/TDX 44 0.03 2034 0.01 0.03 2014-01-03 TSX/TDX 33 0.02 2819 0.01 0.02 2014-01-04 TSX/TDX 55 0.04 2128 0.00 0.04 2014-01-04 TSX/TDX 33 0.05 1939 0.01 0.05 2014-01-09 TSX/TDX 22 0.03 2828 0.01 0.03 2014-01-10 TSX/TDX 44 0.03 2083 0.01 0.03 2014-01-10 TSX/TDX 22 0.10 2104 0.01 0.10 2014-01-14 TSX/TDX 44 0.01 3685 0.01 0.01 2014-01-15 TSX/TDX 33 0.05 2236 0.01 0.05 2014-01-19 TSX/TDX 33 0.02 3652 0.01 0.02 2014-01-31 TSX/TDX 22 0.03 2647 0.01 0.03 2014-02-27 TSX/TDX 44 0.03 3163 0.01 0.03 2014-02-28 TSX/TDX 55 0.05 2235 0.00 0.05 2014-03-24 TSX/TDX 11 0.08 1958 0.02 0.08 2014-03-27 TSX/TDX 11 0.03 15610 0.02 0.03 2014-04-04 TSX/TDX 33 0.04 1921 0.01 0.04 2014-04-10 TSX/TDX 22 0.05 1895 0.01 0.05 2014-07-25 TSX/TDX 11 0.07 1184 0.02 0.08 2014-08-05 TSX/TDX 33 0.05 1130 0.01 0.05 2014-08-06 TSX/TDX 22 0.03 2495 0.01 0.03 2014-08-11 TSX/TDX 33 0.02 2649 0.01 0.02 2014-08-11 TSX/TDX 22 0.08 1340 0.01 0.08 2014-08-22 TSX/TDX 11 0.08 3049 0.02 0.08 2014-08-27 TSX/TDX 11 0.08 1215 0.02 0.09 2014-12-16 TSX/TDX 11 0.03 15265 0.02 0.03

datasets Mean values: 382 All 0.07 11717 0.05 0.08 59 ERS 0.14 26475 0.04 0.15 5 R1 0.09 32794 0.06 0.11 41 ENVISAT 0.11 30240 0.04 0.12 43 ALOS 0.05 13868 0.01 0.05 234 TSX/TDX 0.06 4414 0.01 0.06

129 Table 13-6: Observed parameters of the individual glaciers. lf – length of ice front; A – Glacier area in the respective period; dA – Change in glacier area between 1985 and 2015; Area change category – see definition in Section 4.2 in the main manuscript; Date vs - date of first velocity measurement; Date vE – date of last velocity measurement; dt - mean time period of velocity measurements, vs – mean of earliest velocity measurements (1992-1996); vE – mean of latest velocity measurements (2010-2014), dv – mean velocity change; nv – sum of velocity measurements in the observation period (dt); Velocity change category – see definition in Table 3 in the main manuscript; FS – Ice mass flux at the time of earliest velocity measurements (1992-1996); FE – Ice mass flux at the time of latest velocity measurements (2010-2014); ti – mean ice thickness at the flux gates, hmax – average maximum altitude of individual basins, HI – Hypsometric Index of the basin; Hypsometric category – see Table 4 in the manuscript; FA – Flux gate to catchment size ratio; Group – Classification of glaciers in sector “West” according to hierarchical cluster analysis in sector 4.6 in the main manuscript. Area change Date v Date v Vel. Change Hypsometric Sector Basin l [m] A [km²] A [km²] dA [km²] S E dt [a] v [m/d] v [m/d] dv [m/d] dv [%] n F [Gt/a] F [Gt/a] t [m] h [m] HI FA Group f 1985-1990 2010-2015 category [yyyy-mm-dd] [yyyy-mm-dd] S E v Category S E i max Category East ADD ID: 2707 5535 28.78 26.82 -1.96 retreated 1994-02-01 2013-12-24 19.91 0.553 0.107 -0.446 -80.690 35 decreased 0.04±0.03 0.01±0.01 112.18±50.92 1278 5.14very bottom-heavy 0.0056 ADD ID: 2731 10955 56.92 55.85 -1.06 retreated 1995-12-18 2010-12-31 15.05 0.358 0.093 -0.265 -73.985 12 decreased 0.06±0.03 0.02±0.01 266.75±123.06 1327 2.93very bottom-heavy 0.0055 Aitkenhead 6532 156.70 155.11 -1.59 retreated 1995-12-18 2014-03-27 18.28 0.108 0.147 0.039 36.646 46 peak 0.18±0.08 0.26±0.12 202.47±83.68 1746 -1.23 top-heavy 0.0024 Broad Valley 5948 246.73 246.08 -0.64 retreated 1994-02-01 2010-12-31 16.92 0.743 0.230 -0.512 -68.969 10 stable 0.98±0.07 0.32±0.02 214.06±20.51 1118 -1.02 equidimensional 0.0005 Diplock 8916 235.30 234.14 -1.16 retreated 1995-12-18 2014-12-16 19.01 0.559 0.618 0.059 10.626 33 trough 0.29±0.06 0.34±0.07 257.01±50.49 1845 -1.44 top-heavy 0.0017 Eyrie 6570 89.53 84.35 -5.18 retreated 1992-12-25 2010-12-31 18.03 0.865 0.169 -0.696 -80.499 7 decreased 0.72±0.17 0.15±0.03 273.76±59.90 1076 2.39very bottom-heavy 0.0035 Russell East 2156 93.75 93.38 -0.37 retreated 1992-12-25 2013-12-07 20.96 0.963 0.389 -0.573 -59.559 34 decreased 0.26±0.13 0.11±0.05 305.52±139.81 1370 1.48 bottom-heavy 0.0035 TPE10 5465 225.96 225.24 -0.72 retreated 1994-02-01 2010-12-31 16.92 0.633 0.102 -0.531 -83.928 9 peak 0.44±0.94 0.07±0.16 311.70±142.30 1386 1.43 bottom-heavy 0.0033 TPE130 4493 40.58 38.72 -1.86 retreated 1994-02-01 2013-12-24 19.91 0.244 0.201 -0.043 -17.777 36 peak 0.00±0.04 0.00±0.04 189.96±86.00 983 2.07very bottom-heavy 0.0076 TPE31 11684 52.70 48.76 -3.94 retreated 1992-12-25 2014-12-16 21.99 1.844 0.344 -1.500 -81.352 25 decreased 0.72±0.39 0.15±0.08 140.33±63.51 1490 3.50very bottom-heavy 0.0076 TPE32 4071 108.63 108.24 -0.38 retreated 1992-12-25 2014-03-27 21.27 1.549 0.755 -0.794 -51.271 36 decreased 0.47±0.22 0.24±0.11 301.92±139.32 1646 1.46 bottom-heavy 0.0037 TPE34 2814 22.91 22.25 -0.66 retreated 1992-12-25 2010-12-31 18.03 1.076 0.076 -1.000 -92.937 10 decreased 0.05±0.02 0.00±0.00 150.25±67.75 500 -1.37 top-heavy 0.0023 Victory 9975 180.30 178.75 -1.55 retreated 1995-12-18 2013-12-24 18.03 3.448 0.765 -2.683 -77.809 26 trough 1.38±0.73 0.32±0.17 395.54±180.40 1645 2.11very bottom-heavy 0.0041 mean 18.79 0.995 0.307 -0.688 -69.121 1339 sum 85114 1538.78 1517.71 -21.07 319 5.59±2.91 1.99±0.87 240±92 East-Ice-Shelf ADD ID: 2558 5890 60.2432737 56.31 -3.94 retreated 1993-01-29 2010-12-29 17.93 0.435 0.353 -0.082 -18.758 30 peak 0.16±0.07 0.10±0.04 271.13±112.99 1840 9.08very bottom-heavy 0.0067 ADD ID: 2668 20996 162.3242515 160.93 -1.39 retreated 1996-02-13 2014-12-16 18.85 0.435 0.340 -0.095 -21.821 27 peak 0.42±0.23 0.28±0.16 184.29±42.46 1342 2.88very bottom-heavy 0.0041 APPE 31872 696.24 639.85 -56.39 retreated 1993-01-12 2014-12-16 21.94 0.869 0.853 -0.015 -1.766 114 fluctuating 0.65±0.05 0.66±0.05 209.69±14.34 1964 1.82very bottom-heavy 0.0003 ArronIcefall 10557 152.356315 131.88 -20.48 retreated 1993-01-12 2011-01-22 18.04 0.532 0.288 -0.244 -45.793 39 peak 0.15±0.08 0.06±0.03 299.09±95.94 1979 -1.08 equidimensional 0.0061 Boydell 1954 108.03917516 94.95 -13.09 retreated 1995-12-18 2014-12-16 19.01 0.290 0.975 0.685 236.007 38 peak 0.03±0.01 0.09±0.02 330.08±78.88 1842 -1.07 equidimensional 0.0009 DBE 12140 658.91 627.24 -31.67 retreated 1993-01-12 2014-02-27 21.14 0.535 0.950 0.415 77.569 88 peak 0.67±0.16 0.86±0.20 279.86±71.12 2167 1.37 bottom-heavy 0.0011 Drygalski 14018 990.41 964.49 -25.92 retreated 1993-01-29 2010-12-29 17.93 1.422 1.641 0.219 15.374 29 peak 1.51±0.09 1.34±0.08 241.04±32.55 2043 1.60very bottom-heavy 0.0003 LAB2 4157 38.388905 37.47 -0.92 retreated 1993-01-29 2010-12-29 17.93 0.060 0.065 0.006 9.726 17 peak 0.01±0.01 0.01±0.01 302.41±109.06 1779 3.76very bottom-heavy 0.0046 LAB32 5534 66.3815992 63.60 -2.78 retreated 1993-01-12 2010-12-29 17.97 0.221 0.284 0.063 28.300 19 stable 0.05±0.02 0.05±0.03 294.13±130.74 1841 3.21very bottom-heavy 0.0046 Sjögren 3838 329.2977714 300.73 -28.57 retreated 1992-12-25 2014-12-16 21.99 0.570 0.638 0.068 11.897 42 peak 0.50±0.15 0.45±0.13 312.65±90.34 1926 1.97very bottom-heavy 0.0014 TPE114 7310 126.38516106 110.61 -15.78 retreated 1996-02-29 2014-12-16 18.81 0.098 0.183 0.084 85.924 56 stable 0.10±0.02 0.13±0.02 174.63±25.11 1759 2.96very bottom-heavy 0.0014 TPE61 2943 54.341324 49.09 -5.25 retreated 1993-01-12 2011-01-22 18.04 0.406 0.276 -0.130 -31.942 42 peak 0.10±0.02 0.05±0.01 223.50±49.90 1981 2.78very bottom-heavy 0.0022 TPE62 6700 211.810704 209.40 -2.41 retreated 1992-12-25 2011-01-22 18.09 0.372 0.448 0.076 20.424 43 peak 0.21±0.04 0.21±0.04 299.84±52.70 2118 2.43very bottom-heavy 0.0013 mean 19.05 0.480 0.561 0.081 16.811 1891 sum 127909 3655.13 3446.54 -208.59 584 4.55±0.95 4.29±0.83 263±70 West AMR 7773 137.24 136.73 -0.51 retreated 1993-02-01 2014-08-22 21.57 0.157 0.837 0.679 431.515 21 increased 0.11±0.03 0.54±0.15 146.09±38.02 1884 -3.82 very top-heavy 0.0021 1 Andrew 2951 47.05 44.41 -2.64 retreated 1992-12-25 2014-08-27 21.68 0.453 0.358 -0.095 -21.030 119 decreased 0.23±0.09 0.17±0.06 149.31±52.93 1731 1.99very bottom-heavy 0.0057 4 Bagshawe-Grubb 10720 280.43 280.17 -0.26 stable 1993-02-01 2010-12-22 17.90 0.302 0.233 -0.069 -22.782 14 stable 0.67±0.26 0.49±0.19 169.59±50.97 2169 -2.88 very top-heavy 0.0019 1 Bayly 4149 47.89 47.32 -0.57 retreated 1993-02-01 2014-08-22 21.57 0.419 0.912 0.493 117.584 42 increased 0.07±0.02 0.15±0.04 145.48±35.79 1529 -1.06 equidimensional 0.0027 2 Blanchard 2005 38.00 37.63 -0.36 retreated 1993-02-01 2014-08-22 21.57 0.341 1.084 0.744 218.153 31 increased 0.03±0.01 0.10±0.03 185.11±50.68 2060 1.53very bottom-heavy 0.0025 2 Bleriot 8527 182.20 180.69 -1.50 retreated 1993-02-01 2014-04-10 21.20 0.836 0.300 -0.536 -64.134 30 decreased 1.59±0.34 0.54±0.12 192.75±36.76 1943 1.28 bottom-heavy 0.0019 3 CLM 12682 809.85 809.58 -0.27 stable 1993-02-01 2010-12-29 17.92 0.388 0.396 0.008 2.157 34 peak 0.52±0.25 0.52±0.25 288.99±131.69 2191 1.13 equidimensional 0.0016 2 Deville 8699 34.99 34.79 -0.20 stable 1996-02-15 2010-12-22 14.86 0.364 0.127 -0.237 -65.116 12 decreased 0.15±0.02 0.05±0.01 148.52±19.42 1389 -1.19 equidimensional 0.0025 3 DGC10 6423 23.47 23.40 -0.06 stable 1993-02-01 2014-04-10 21.20 0.116 0.580 0.465 401.477 24 increased 0.02±0.01 0.10±0.04 126.95±52.81 1219 -1.10 equidimensional 0.0064 2 DGC13 1950 10.95 10.76 -0.18 retreated 1996-02-15 2014-04-10 18.16 0.285 0.205 -0.081 -28.256 27 peak 0.02±0.01 0.02±0.01 124.97±49.72 901 1.28 bottom-heavy 0.0071 3 DGC14 1684 5.66 5.64 -0.02 stable 1996-02-15 2014-04-10 18.16 0.096 0.113 0.018 18.626 25 stable 0.00±0.00 0.00±0.00 136.67±59.62 884 1.90very bottom-heavy 0.0109 3 DGC22 2188 8.98 9.10 0.12 stable 1996-02-15 2014-04-10 18.16 0.190 0.084 -0.106 -55.993 31 stable 0.04±0.02 0.02±0.01 76.64±25.70 1113 -1.24 top-heavy 0.0148 3 DGC23 1868 15.92 15.91 0.00 stable 1993-02-01 2014-08-22 21.57 0.414 1.025 0.611 147.314 37 increased 0.02±0.01 0.05±0.01 107.90±24.90 1379 -1.33 top-heavy 0.0023 2 DGC25 2693 14.12 14.27 0.15 stable 1993-02-01 2014-08-22 21.57 0.363 0.820 0.457 125.807 37 increased 0.02±0.01 0.04±0.01 147.60±33.92 1850 1.52very bottom-heavy 0.0028 2 DGC31 1466 13.30 13.06 -0.24 retreated 1996-02-15 2010-12-11 14.83 0.132 0.204 0.072 54.579 10 stable 0.01±0.00 0.01±0.00 88.04±29.81 1488 1.86very bottom-heavy 0.0029 2 DGC39 1331 15.07 14.97 -0.10 retreated 1993-02-01 2010-12-22 17.90 0.529 0.164 -0.365 -69.044 8 decreased 0.02±0.01 0.01±0.00 131.43±59.52 1472 1.02 equidimensional 0.0040 3 DGC72 4990 38.39 38.09 -0.30 stable 1993-02-01 2014-04-10 21.20 0.359 0.695 0.336 93.651 19 peak 0.12±0.03 0.22±0.05 117.16±27.95 1706 1.17 equidimensional 0.0027 2 DGC8 3340 9.34 8.91 -0.43 retreated 1993-02-01 2014-04-10 21.20 0.177 0.241 0.064 36.012 32 stable 0.02±0.01 0.02±0.01 143.81±47.56 1061 2.07very bottom-heavy 0.0094 4 Krebs 3152 34.80 35.27 0.47 advanced 1993-02-01 2014-04-10 21.20 0.866 0.738 -0.128 -14.780 16 peak 0.13±0.01 0.10±0.01 197.17±9.47 2029 -2.00 very top-heavy 0.0006 1 Landau 2330 33.99 33.90 -0.08 stable 1996-02-13 2014-08-27 18.55 0.069 0.727 0.658 954.866 49 increased 0.01±0.00 0.12±0.03 165.61±41.07 1747 -1.79 very top-heavy 0.0027 1 Leonardo 3632 84.22 83.72 -0.49 retreated 1993-02-01 2014-08-22 21.57 0.281 1.493 1.212 431.732 24 increased 0.05±0.01 0.27±0.05 194.06±32.92 2106 1.06 equidimensional 0.0009 2 Mc Neile 2507 184.56 184.66 0.10 stable 1995-12-19 2014-08-27 18.70 0.207 0.699 0.492 237.738 30 increased 0.18±0.03 0.58±0.11 230.72±35.57 1882 -4.58 very top-heavy 0.0006 1 Montgolfier 4486 55.20 55.06 -0.13 stable 1993-02-01 2014-08-22 21.57 0.141 1.371 1.230 872.806 21 increased 0.03±0.01 0.27±0.08 147.83±41.38 1929 -1.32 top-heavy 0.0022 1 Nobile 2361 57.04 56.78 -0.26 retreated 1993-02-01 2014-04-10 21.20 0.233 0.372 0.139 59.586 13 peak 0.06±0.02 0.09±0.03 218.35±63.82 1901 -1.28 top-heavy 0.0018 1 Orel 5399 19.02 18.11 -0.92 retreated 1996-02-15 2010-12-22 14.86 0.229 0.172 -0.057 -25.010 9 stable 0.04±0.01 0.03±0.01 134.99±35.25 1148 1.95very bottom-heavy 0.0066 4 Pettus-GavinIce 3535 330.88 330.67 -0.21 stable 1992-12-25 2014-08-05 21.62 0.686 0.385 -0.301 -43.827 33 peak 1.10±0.63 0.60±0.34 298.52±136.81 1846 1.24 bottom-heavy 0.0030 2 Renard 5904 118.15 117.24 -0.91 retreated 1993-02-01 2014-08-22 21.57 0.212 1.698 1.486 699.238 36 increased 0.07±0.01 0.51±0.08 147.83±21.87 2043 -1.82 very top-heavy 0.0011 1 Rozier 5984 35.57 35.07 -0.50 retreated 1996-02-15 2014-08-22 18.53 0.977 0.944 -0.033 -3.420 41 peak 0.12±0.05 0.11±0.04 186.23±66.13 2061 2.70very bottom-heavy 0.0036 2 Russell West 3450 329.28 328.95 -0.33 retreated 1996-02-29 2014-08-27 18.50 1.072 1.759 0.687 64.111 18 increased 0.74±0.34 1.19±0.55 358.92±163.91 1645 1.44 bottom-heavy 0.0028 2 Sabine 1795 83.09 82.78 -0.31 retreated 1993-02-01 2014-08-27 21.58 0.239 0.348 0.109 45.520 82 increased 0.16±0.08 0.22±0.11 270.00±116.73 1843 1.21 bottom-heavy 0.0070 2 SBG 10917 327.95 327.75 -0.20 stable 1993-02-01 2010-12-29 17.92 0.298 0.306 0.007 2.395 35 peak 0.72±0.34 0.72±0.34 294.02±134.20 2220 1.08 equidimensional 0.0047 2 Stringfellow-Henson 7775 670.38 669.74 -0.64 retreated 1993-02-01 2014-02-28 21.09 1.100 1.233 0.132 12.029 24 fluctuating 0.56±0.35 0.62±0.38 496.51±227.37 2167 1.55very bottom-heavy 0.0026 2 Temple 12056 453.96 453.22 -0.74 retreated 1992-12-25 2014-08-11 21.64 1.544 1.516 -0.028 -1.821 92 fluctuating 1.20±0.46 1.14±0.44 345.48±129.80 1962 -1.06 equidimensional 0.0031 1 TPE11 1947 70.06 70.13 0.07 stable 1995-12-20 2013-12-24 18.02 0.184 1.203 1.018 552.655 21 increased 0.02±0.01 0.15±0.07 182.27±82.61 1268 1.05 equidimensional 0.0028 2 TPE125 8741 40.41 40.13 -0.27 stable 1992-12-25 2013-12-24 21.01 0.415 0.260 -0.155 -37.319 23 fluctuating 0.07±0.08 0.04±0.05 142.93±64.63 1104 1.82very bottom-heavy 0.0116 3 TPE126 16295 145.52 147.80 2.28 advanced 1995-12-19 2014-08-27 18.70 0.287 0.306 0.019 6.542 58 peak 0.18±0.11 0.18±0.11 164.27±67.88 1655 2.20very bottom-heavy 0.0060 2 TPE39 9931 139.49 139.40 -0.08 stable 1995-12-19 2013-12-07 17.98 0.341 0.690 0.348 102.092 21 peak 0.02±0.04 0.03±0.09 233.03±106.71 1384 1.13 equidimensional 0.0051 2 TPE40 13405 184.11 184.69 0.58 stable 1992-12-25 2013-12-24 21.01 0.718 0.406 -0.312 -43.414 27 decreased 0.48±0.24 0.26±0.13 246.46±112.76 1386 1.01 equidimensional 0.0059 3 TPE41 9256 53.13 53.24 0.11 stable 1995-12-19 2013-12-07 17.98 0.326 0.281 -0.046 -13.987 26 stable 0.08±0.07 0.07±0.05 156.80±70.96 1094 1.98very bottom-heavy 0.0107 3 TPE46 2785 33.94 34.34 0.41 advanced 1992-12-25 2014-08-27 21.68 0.935 0.881 -0.054 -5.756 43 fluctuating 0.12±0.03 0.11±0.03 225.61±48.05 1843 -1.86 very top-heavy 0.0026 1 TPE50 2987 31.32 31.53 0.21 advanced 1992-12-25 2014-08-27 21.68 0.450 0.517 0.067 14.899 114 peak 0.03±0.01 0.03±0.01 120.07±33.86 1839 1.13 equidimensional 0.0023 2 TPE57 20111 100.43 100.34 -0.10 stable 1993-02-01 2010-12-29 17.92 0.317 0.230 -0.087 -27.382 32 peak 0.22±0.12 0.15±0.08 123.21±55.73 1132 1.31 bottom-heavy 0.0090 3 TPE8 5582 111.74 112.24 0.49 advanced 1996-02-11 2013-12-24 17.88 0.991 0.739 -0.252 -25.395 17 trough 0.27±0.13 0.20±0.09 267.69±122.89 1104 1.19 equidimensional 0.0035 3 TPE9 3735 48.96 49.64 0.68 advanced 1995-12-19 2013-12-24 18.03 0.355 0.150 -0.205 -57.744 20 decreased 0.11±0.06 0.04±0.02 178.13±80.44 1085 1.41 bottom-heavy 0.0057 3 Wellman 3449 48.67 48.48 -0.19 stable 1996-02-15 2014-04-10 18.16 0.161 0.255 0.094 58.300 25 stable 0.07±0.02 0.10±0.03 163.53±41.54 1772 1.47 bottom-heavy 0.0037 2 Wheatstone 4642 52.66 52.18 -0.48 retreated 1993-02-01 2010-12-22 17.90 0.355 0.258 -0.097 -27.262 11 peak 0.06±0.02 0.04±0.01 134.70±40.08 1569 1.21 bottom-heavy 0.0029 2 Whitecloud 3711 177.77 177.66 -0.11 stable 1992-12-25 2014-08-11 21.64 0.454 0.481 0.027 5.848 59 fluctuating 0.59±0.15 0.61±0.16 314.16±71.95 1950 -2.94 very top-heavy 0.0013 1 Woodbury 1464 20.24 20.03 -0.21 retreated 1993-02-01 2014-08-11 21.54 0.155 0.239 0.084 53.784 27 stable 0.00±0.00 0.01±0.00 103.61±34.93 1862 1.02 equidimensional 0.0024 2 mean 20 0.427 0.605 0.177 41.487 1636 sum 268763 5809.33 5800.18 -9.14 1600 11.20±4.56 11.65±4.52 189±65 All Glaciers mean 19 0.537 0.545 0.008 1629 sum 481786 11003.23 10764.42 -238.81 2503 21.94±8.54 17.93±6.22 211±71 Appendix

Figure S1-S13: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "East". 131 Appendix

Figure S14-S26: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "East-Ice-Shelf". 132 Appendix

Figure S27-S41: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "West".

133 Appendix

Figure S45-S56: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "West". 134 Appendix

Figure S57-S71: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "West". 135 Appendix

Figure S72-S74: Temporal trend of surface velocity (red) and area (blue) changes of glaciers in sector "West".

136