Remote Sensing of Environment 139 (2013) 50–59

Contents lists available at ScienceDirect

Remote Sensing of Environment

j ourna l homepage: www.e lsev ie r .com/ locate / rse

Surface velocities and ice-front positions of eight major glaciers in theSouthern Patagonian Ice Field, South America, from 2002 to 2011

Minami Muto, Masato Furuya ⁎Department of Natural History Sciences, Graduate School of Science, Hokkaido University, N10W8, Kita-ku, Sapporo 060-0810, Japan

⁎ Corresponding author. Tel.: +81 11 706 2759.E-mail address: furuya@mail.sci.hokudai.ac.jp (M. Furu

0034-4257/$ – see front matter © 2013 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.rse.2013.07.034

a b s t r a c t

a r t i c l e i n f o

Article history:Received 14 December 2012Received in revised form 28 July 2013Accepted 30 July 2013Available online xxxx

Keywords:Glacier velocitySouthern PatagoniaCalving glacierSynthetic aperture radar

The Patagonian Ice Fields are known to have undergone rapid retreat of frontal positions and significantthinning of its glaciers over the past decades. However, surface velocities have been measured at only afew of these glaciers. Thus, it remains uncertain if and to what extent the glacier dynamics has changedover time and contributed to ice loss in these ice fields. In this study, we examine the temporal evolutionof flow velocities and ice-front positions at eight major glaciers in the Southern Patagonian Ice Field (SPI;Hielo Patagónico Sur) by using Advanced Synthetic Aperture Radar (ASAR) images from the Environmental Sat-ellite (Envisat) launched in 2002by the European Space Agency andAdvanced LandObservation Satellite/PhasedArray-type L-band Synthetic Aperture Radar (ALOS/PALSAR) data recorded from 2002 to 2011. The examinedeight glaciers include Glaciar Jorge Montt, Occidental, Pio XI (or Brüggen), O'Higgins, Viedma, Upsala, PeritoMoreno, and Grey. Not all the glaciers revealed significant changes in frontal positions and flow velocitiesin the study period. We detected significant temporal velocity changes at Glaciar Upsala, Jorge Montt,Occidental, and Pio XI. Among these four glaciers, Glaciars Upsala, Jorge Montt, and Occidental revealed sig-nificant acceleration and terminus retreat andwere undergoing dynamic-thinning. Themarkedly different abso-lute velocities but equally large longitudinal near-terminus stretching at the three glaciers support a calvingmodel based on crevasse-depth criteria, which predict a calving position where crevasse-depths are equal toice thickness; crevasse-depths are controlled by the longitudinal stretching rate. Meanwhile, Glaciar Pio XI re-vealed complex spatial and temporal evolution in surface velocities without significant retreat, and its dynamicsremains enigmatic.

© 2013 Elsevier Inc. All rights reserved.

1. Introduction

The Southern Patagonian Ice Field (SPI; Hielo Patagónico Sur) con-tains the largest temperate glaciers in southern hemisphere and coversapproximately 13,000 km2 area from approximately 48.5°S to 51.5°S(Aniya, Sato, Naruse, Skvarca, & Casassa, 1996; Fig. 1). The SPI consistsof 48 outlet glaciers with an average altitude of 1355 m above sealevel. All these outlet glaciers, except for two, are calving (Aniya et al.,1996). The ones on the western Chilean side mostly calve into fjordsas tidewater calving glaciers, whereas those on the eastern Argentinaside terminate into proglacial lakes (Aniya et al., 1996; Warren &Aniya, 1999). Their retreat in the frontal positions and rapid thinningtoward the end of the 20th century has been reported, suggesting asignificant contribution to global sea level increases (Aniya, Sato,Naruse, Skvarca, & Casassa, 1997; Rignot, Rivera, & Casassa, 2003).Moreover, Gravity Recovery And Climate Experiment (GRACE)observation data recorded since 2002 indicate a total mass loss ratefrom both the Northern Patagonian Ice Field and the SPI to be rangingfrom –23.0 ± 9.0 Gt a−1 to –26.0 ± 6.0 Gt a−1 (Chen, Wilson, Tapley,

ya).

ghts reserved.

Blankenship, & Ivins, 2007; Ivins et al., 2011; Jacob, Wahr, Pfeffer, &Swenson, 2012). GRACE-based mass-loss estimates in the beginningof the 21st century are larger than the corresponding volume loss ratederived in the late 20th century (Rignot et al., 2003). The recentlyreported surface height change rates, based on a time series of digitalelevation models (DEMs), are also consistent with GRACE-basedestimates, which indicates an even more rapid drawdown over thepast decade (Willis, Melkonian, Pritchard, & Rivera, 2012).

While near-surface temperature increases associated with globalwarming presumably contribute to ice loss due to melting, the acceler-ated ice losses reported on Greenland and Antarctica have been linkedto changes in glacier dynamics as well (Pritchard, Arthern, Vaughan, &Edwards, 2009; Rignot & Kanagaratnam, 2006; van den Broeke et al.,2009). In particular, many polar glaciers that calve into the ocean haveundergone significant acceleration, known as dynamic thinning, andhave thus further contributed to ice loss. However, recent studiesdemonstrate spatial–temporal complexities of the outlet glacier veloci-ties in Greenland (Bevan, Murray, Luckman, Hanna, & Huybrechts,2012; Moon, Joughin, Smith, & Howat, 2012). In contrast to theselarge polar glaciers, no detailed glacier velocity maps are available forthe Patagonian Ice Fields with the exception of those for certain accessi-ble glaciers (Naruse, Skvarca, Kadota, & Koizumi, 1992; Naruse, Skvarca,

51M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

Satow, Takeuchi, & Nishida, 1995; Rivera, Corripio, Bravo, & Cisternas,2012; Stueffer, Rott, & Skvarca, 2007; Sugiyama et al., 2011). Satellite-derived velocity maps are also limited to Glaciars Perito Moreno(Ciappa, Pietranera, & Battazza, 2010; Michel & Rignot, 1999; Rott,Stuefer, Siegel, Skvarca, & Eckstaller, 1998) and Glaciar Upsala

74°W 73°W

51°S

50°S

49°S

Pio XI

Occidental

Jorge Montt

O’Higgins

Viedma

Upsala

Perito Moreno

GreyGrey

SouthAmerica

0 1000 2000 3000

m

Fig. 1. Elevation map of the studied area and the glaciers in the South Patagonian Icefield;the elevation data set was obtained NASA's Shuttle Radar Topography Mission (SRTM).Red triangles represent the termini of the eight analyzed glaciers. Blue and light-blueindicate fjord and proglacial lakes, respectively. Glacier outlines were determined on thebasis of the Global Land IceMeasurements from Space (GLIMS) dataset, which is availablethrough the US National Snow and Ice Data Center (Dickmann, 2007; Delgado, 2009;Masiokas, 2009, 2010; Davies, 2012). (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)

(Floricioiu, Eineder, Rott, & Nagler, 2008; Floricioiu, Eineder, Rott,Yague-Martinez, & Nagler, 2009; Sakakibara, Sugiyama, Sawagaki,Marinsek, & Skvarca, 2013; Skvarca, Raup, & De Angelis, 2003).Thus, it remains uncertain if and to what extent dynamic thinninghas affected the Patagonian Ice Fields. In this study, we show the spa-tial and temporal changes in flow velocities at eight major glaciers inthe SPI by applying an offset tracking technique to the intensityimages of the Environmental Satellite/Advanced Synthetic ApertureRadar (Envisat/ASAR) and Advanced Land Observation Satellite/Phased Array-type L-band Synthetic Aperture Radar (ALOS/PALSAR)data recorded from 2002 to 2011. In addition, we used the cloud-freecharacteristics of SAR imagery for examining the terminus changesin order to determine whether the ice acceleration is associatedwith calving episodes. The examined eight glaciers include GlaciarsUpsala, Jorge Montt, Occidental, Pio XI (or Brüggen), O'Higgins,Viedma, Perito Moreno, and Grey (Fig. 1). We selected these glaciersnot only because their sizes were sufficient to be imaged by thespatial resolution of presently available SAR data but also becausethey were more frequently imaged than others so that we could in-crease the temporal resolution, which allowed us to detect rapidchanges in the ice dynamics.

As in the case of Greenland (Moon et al., 2012), we subsequentlyshow that not all of the glaciers reveal ice acceleration and rapid ter-minus retreat. Submarine melting has been suggested as a principaltriggering mechanism for the accelerated ice motion in Greenland,because the timing of the glacier's acceleration coincided with thatof the ocean water warming near Greenland (Holland, Thomas, deYoung, Ribergaard, & Lyberth, 2008; Rignot, Koppes, & Velicogna,2010; Straneo et al., 2011). The rapid retreat and flow accelerationin the SPI, however, occur at both marine and fresh-water terminat-ing glaciers, which suggests the presence of additional generalprocesses independent of water salinity. While the physically basedcalving model remains elusive (Benn, Warren, & Mottram, 2007),we interpret the observed data sets on the basis of the calvingmodel based on crevasse-depth criteria (Benn, Hulton, & Mottram,2007; Benn, Warren, et al., 2007; Nick, van der Veen, Vieli, & Benn,2010).

2. Data and analysis method

2.1. Satellite data

To generate glacier surface velocity maps, we processed thePALSAR images with a wavelength of 23.6 cm recorded from June2007 to February 2011 from ALOS, which was launched in 2006 bythe Japan Aerospace Exploration Agency (JAXA) (Table 1, Fig. 1). Toextend the analysis period further back to 2003, we also used theC-band (wavelength of 5.6 cm) ASAR images from Envisat launchedin 2002 by the European Space Agency (Table 1, Fig. 1), which oftencontains de-correlation problems; the advantage of the L-band overC-band was shown before (Rignot, 2008; Strozzi et al., 2008). Onthe basis of our observations of insignificant temporal velocitychanges at Glaciar Perito Moreno (Supplementary material), the dif-ferences in penetration depths due to different wavelengths do notappear to affect the inferred velocities. This result is theoreticallyreasonable because the most significant changes in glacier-flowvelocity profile are expected to occur near the glacier bed ratherthan near the surface (Cuffey & Paterson, 2010).

The off-nadir beam angle of PALSAR is 34.3°, which forms ~39° inci-dent angle at theflat ground in the image center. There are two availableimaging modes, fine beam single polarization (FBS, HH) and fine beamdual polarization (FBD; HH and HV).We used data from only HH polar-ization. The difference between FBS and FBD is the slant range (acrosstrack) resolution, which is ~4.7 m for the FBS mode and ~9.4 m forthe FBD mode; the azimuth resolution is 3.1 m regardless of themode. The FBD data are oversampled twice along the range axis. On

Table 1The processed SAR images.

Sensor Path (track) Frame Acquisition date Mode A/D

PALSAR 129 6150 20100216 FBS A129 6150,6160,6170,6180 20100519 FBD A129 6150,6160,6170,6180 20100704 FBD A129 6150,6160,6170,6180 20110104 FBS A129 6150,6160,6170,6180 20110219 FBS A130 6150,6160,6170,6180 20100605 FBD A130 6150,6160,6170,6180 20100721 FBD A130 6150 20101021 FBD A130 6150 20110121 FBS A410 4650 20070619 FBD D410 4650 20070919 FBD D410 4650 20071104 FBD D410 4650 20071220 FBS D411 4650 20070706 FBD D411 4650 20071121 FBD D411 4650 20080106 FBS D414 4600,4620 20070826 FBD D414 4600,4620 20071011 FBD D414 4600,4620 20080111 FBS D414 4600,4620 20080226 FBS D

ASAR 153 4599,4617,4635 20021220 I2 D153 4599,4617,4635 20030228 I2 D153 4599,4617,4635 20030613 I2 D153 4599,4617,4635 20030718 I2 D153 4599,4617,4635 20030926 I2 D153 4599,4617,4635 20031205 I2 D153 4599,4617,4635 20040109 I2 D153 4599,4617,4635 20040910 I2 D153 4599,4617,4635 20050722 I2 D153 4599,4617,4635 20050826 I2 D153 4599,4617,4635 20080815 I2 D196 4599 20021223 I2 D196 4599 20030616 I2 D196 4599 20031208 I2 D196 4599 20040122 I2 D196 4599 20040322 I2 D196 4599 20040426 I2 D196 4599 20040531 I2 D196 4599 20040913 I2 D196 4599 20050620 I2 D196 4599 20050829 I2 D196 4599 20051003 I2 D196 4599 20070730 I2 D376 6147 20050703 I6 A376 6147 20050807 I6 A376 6147 20050911 I6 A382 4635 20030629 I2 D382 4635 20030803 I2 D382 4635 20040125 I2 D382 4635 20040229 I2 D382 4635 20040404 I2 D382 4635 20040509 I2 D382 4635 20040613 I2 D425 4599,4617 20030108 I2 D425 4599,4617 20030212 I2 D425 4599,4617 20030423 I2 D425 4599,4617 20030702 I2 D425 4599,4617 20031015 I2 D425 4599,4617 20040407 I2 D425 4599,4617 20040512 I2 D425 4599,4617 20040825 I2 D425 4599,4617 20080903 I2 D

52 M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

the other hand, the ASAR data used in this study were obtained along adescending path with a local incidence beam angle of ~23°; the ASARdata from an ascending path of the mode IS6 were used only for theanalysis of Glaciar Perito Moreno (Table 1, Supplementary material).This smaller incidence angle lowers the sensitivity to the east–westcomponent of displacement, although it can increase the sensitivity ofthe vertical component. The spatial resolution of ASAR in the slantrange and azimuth direction is 7.8 m and 4.1 m, respectively. We

processed the PALSAR level 1.0 products and ASAR level 0 products togenerate single look complex images.

2.2. Pixel-offset tracking

Although the other approaches are available, such as radar interfer-ometry (Goldstein, Engelhardt, Kamb, & Frolich, 1993) and multipleaperture interferometry (Gourmelen et al., 2011), they are based onphase images thatwill encounter coherence loss and phase unwrappingproblems to quantify larger displacements. In this study, pixel-offsettracking (intensity tracking or speckle tracking) algorithms, based onmaximizing the cross-correlation of the radar image patches, wereused to observe surface velocities (Michel, Avouac, & Taboury, 1999;Strozzi, Luckman, Murray, Wegmüller, & Werner, 2002; Yasuda &Furuya, 2013). We used the intensity tracking algorithm because it ismost suitable for detecting rapidly-flowing glacier velocities with longdata acquisition intervals (Strozzi et al., 2002, 2008). Although weexperimented with a variety of patch sizes and sampling intervals,we eventually deemed a search patch of ~500 m (range) × 600 m(azimuth) area with a sampling interval of ~90 m × 120 m as opti-mum parameters. We set a signal-to-noise ratio (SNR) threshold of4.0; patches below this level were assigned to the missing data.Lower SNR data occur because of the large spatial separation length(Bperp) of the repeating orbits or temporal changes in the surfacescattering characteristics; Bperp denotes the perpendicular compo-nent of the baseline projected onto the radar line of sight (LOS).Unfortunately, the Bperp for the PALSAR data pairs in summer wasrelatively long; however, the operational orbit determination accu-racy is now approaching sub-meter level with standard errors oftens of centimeters or less (Andersen, Aksnes, & Skonnord, 1998;Katagiri & Yamamoto, 2008; Scharroo & Visser, 1998). Moreover,because higher precipitation levels and temperatures occur in sum-mer, we can consider that snowfall and surface melting lower thecross-correlation between the temporally separated image patches,which prevents derivation of velocity data, particularly in theaccumulation zones. We did not mask the non-glacier areas inoffset-tracking, because it was important to verify that non-glacierareas could indeed be estimated to be non-deforming, and thatthere were no artifacts due to topography. The no-displacementsignals outside the glaciated areas shown below validated the sur-face velocity data detected along the glacier itself. Over the glaciatedareas, however, we occasionally encountered spurious signals thatindicated, for instance, extraordinary fast velocities or physicallyunrealistic discontinuities in the velocity distributions. We consid-ered that these spurious data were caused by erroneous matchingof objects that actually differed; thus, we visually eliminated suchnoise.

Because the satellite does not repeat exactly the same paths, theeffect of fore-shortening also differs in each path, which results inan artifact offset over rugged terrain (Kobayashi, Takada, Furuya, &Murakami, 2009; Michel et al., 1999; Sansosti, Berardino, Manunta,Serafino, & Fornaro, 2006). We reduced this artifact by applying anelevation-dependent co-registration by employing NASA's ShuttleRadar Topography Mission (SRTM-4) DEM data with a 3-arcsecresolution, in which the gaps in the original SRTM data were filled(Jarvis, Reuter, Nelson, & Guevara, 2008). Although the SRTM datamay include inherent errors and are nearly a decade old, we usedthem because they are free of gaps and apparent noise. Because notopography-correlated offsets were apparent, we considered the ar-tifact offset to be adequately corrected; it is unlikely that the errorsin the SRTM DEM influenced our velocity data.

The pixel-offset tracking technique provides range and azimuthoffsets, both of which are linear combinations of the three-dimensional(3D) displacements. The range offset, Ura, is a projection of the 3Dsurface displacements onto the slant radar LOS direction, while theazimuth offset, Uaz, is a projection of the 3D surface displacements

53M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

along the satellite-track direction. Using the satellite's heading angle Hmeasured clockwise from the north and the microwave's incidenceangle to the ground I, each offset can be represented as follows:

UraUaz

� �¼ cos H sin I − sin H sin I − cos I

sin H cos H 0

� � UeUnUz

0@

1A; ð1Þ

where Ue, Un, and Uz are defined as eastward, northward, and upwardpositive displacements, respectively. The imaging angles and expres-sions of Ura and Uaz for each satellite's observation mode are indicatedin Table 2.

Because the two displacement maps do not allow us to resolve the3D displacements, we derived the surface velocity data by using theparallel flow assumption (Joughin, Kwok, & Fahnestock, 1996). Al-though either the range or azimuth offset data allow us to derivethe surface velocity with the parallel flow assumption, we usedboth data sets to solve the over-determined problem by the least-squares method. The local topographic gradient vector was estimat-ed from the SRTM-4 data. Because only the local flow orientationand slope angles are necessary in the argument of trigonometricfunctions for parallel flow approximation, the ice-thickness changesthat may have occurred between the SRTM and PALSAR/ASAR dataacquisition would be insignificant as potential error sources in thecomputation of a unit vector.

Although the effects of atmospheric propagation and inaccuratesatellite orbit generate significant errors in the InSAR phase data, theseerrors are insignificant in the offset tracking data. This is because the er-rors in the offset tracking data are due to those in the image registration,and are larger than those due to atmospheric propagation and satelliteorbit (Strozzi et al., 2002). The uncertainties of the offset measurementshave been estimated to be ~0.3–0.4 m at the non-deforming ruggedterrain in the ALOS/PALSAR data with a 46-day interval (Kobayashiet al., 2009). If the ice is flowing at a constant rate over the data acqui-sition interval, the error in the velocity estimates can be inferred as0.005–0.01 m/day, and will be inversely proportional to the dataacquisition interval (Strozzi et al., 2002). However, errors in pixel-offset tracking can also originate from other sources.

The most significant error source is the temporal de-correlationbetween image patches due to changes in objects' scattering properties.Thus, all the data pairs shown in this study have the shortest possible46- and 35-day temporal separation, which denote the PALSAR andASAR recurrent periods, respectively. The ASAR datamore often indicat-ed larger errors, probably because the short wavelength data tend toundergo the de-correlation problem. The magnitude of errors alsovaries seasonally from less than 5 m/a up to 30 m/a, depending on theseason of the SAR data acquisition (Yasuda & Furuya, 2013).

For point sites such as those at theGlaciar Upsala and the site C at theGlaciar Jorge Montt, we averaged the velocities at the nearby 500 m2

area and considered the derived standard deviation as the measure-ment error, which ranges from less than 0.1 m/day to ~0.5 m/day. Fortransverse velocity profile, we selected five velocity profiles that areclosely-located within 100 m, and computed the average and standarddeviation along the transverse direction. The velocities and error barsfor the transverse profiles shown belowwere derived by further averag-ing the data along the transverse direction; the velocity profiles and

Table 2The imaging angles and offsets as a function of 3-D displacements.

Sensor Mode (A/D) H (deg.) I (deg.)

PALSAR A −20 39PALSAR D −159 39ASAR IS6 (A) −22 41ASAR IS2 (D) −162 23

errors along the transverse direction are shown in the Supplementarymaterial.

2.3. Ice-front position changes

Cloud-free SAR intensity images permit the tracking of ice-frontpositions over time if clear image contrasts are present between theice andwater surfaces. By using the original single look complex images,the intensity images were derived by multi-looking in range and azi-muth directions. The spatial resolution was 40 m × 40 m. We visuallyidentified the ice-front positions at each SAR intensity image; we didnot performmechanized automatic detection. The front position changeshown below is referenced to the first SAR image in 2002 or 2003, andrepresents the average change derived by dividing the total area ofeach polygon by a fixed width of a reference profile in the upstreamdirection (Moon & Joughin, 2008). Errors in digitized frontal positionsarose mainly from our misidentification of the frontal positions, whichcould be attributed to image resolution and the lower image contrastbetween the ice and water surfaces. Images with unclear contrastsbetween the ice and water surfaces were not used in this study. Bevanet al. (2012) evaluated the accuracies of ±49 m for ERS1/2 and ASARSAR images. We consider that the accuracies of our identified frontalpositions are comparable to or better than the Bevan et al.'s estimate,because the resolution is close to those of ERS1/2 and ASAR in PALSAR'sFBD mode and better in PALSAR's FBS mode.

3. Results

Regarding temporal evolution of surface velocities at the eight largeglaciers examined in this study from 2003 to 2011, Glaciars Upsala,Jorge Montt, Occidental, and Pio XI revealed significant acceleration ofgreater than 30%. Although temporal fluctuations possibly due toseasonalities were recognized, no large velocity changes were evidentat Glaciars Grey, PeritoMoreno, Viedma, andO'Higgins.With the excep-tion of Pio XI, these accelerated glaciers also revealed significantretreating in the frontal positions. In this section, we present the obser-vation results of the Glaciars Upsala, JorgeMontt, Occidental, and Pio XI;the results of the remaining glaciers are shown in the Supplementarymaterial.

3.1. Glaciar Upsala

Glaciar Upsala, the third largest glacier in the SPI, contains three ter-mini, of which the westernmost terminus is the largest and calves intoBrazo Upsala of Lago Argentino (Skvarca et al., 2003). The water depthis 400–500 m (Naruse & Skvarca, 2000). A rapid retreat of the frontalposition since 1978 has been reported (Aniya & Skvarca, 1992; Naruse& Skvarca, 2000; Naruse, Skvarca, & Takeuchi, 1997; Skvarca, Raup, &De Angelis, 2002; Skvarca, Satow, Naruse, & Leiva, 1995). The thinningrate near the glacier front was reported to be 11 m/a between 1990and 1993 (Skvarca et al., 1995) and 10 ± 2 m/a between 2000 and2005 (Willis et al., 2012), but accelerated to 24.8 ± 2.4 m/a between2005 and 2011 (Willis et al., 2012). Annual velocities near the terminusbased on satellite imagery were first derived by Skvarca et al. (2003),who applied a cross-correlationmethod to Landsat 7 Enhanced Thematic

Ura Uaz

0.59Ue + 0.22Un − 0.78Uz −0.34Ue + 0.94Un

−0.59Ue + 0.22Un − 0.78Uz −0.36Ue − 0.93Un

0.61Ue + 0.25Un − 0.75Uz −0.37Ue + 0.93Un

− 0.37Ue + 0.12Un − 0.92Uz −0.31Ue − 0.95Un

Fig. 3. Temporal evolution of the frontal position at Glaciar Upsala between 2002 and2011. Background is a PALSAR-based scattering intensity image acquired on February19, 2011.

54 M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

Mapper Plus (ETM+) images acquired between 2000 and 2001; thisapproach is similar to the currently used intensity tracking approach.Near the fastest-flowing western terminus, Skvarca et al. (2003) derivedaverage velocities of approximately 1600 m/a, equivalent to 4.4 m/day,which were in good agreement with the field measurement obtainedin 1995 (Skvarca et al., 1995). Sakakibara et al. (2013) extended the anal-ysis period to 2011. On the basis of TerraSAR-X data acquired betweenDecember 2007 and January 2008, Floricioiu et al. (2008) reported amaximum velocity of 5.6 m/day, which is significantly faster than theearlier observations in 1995 and 2001; Floricioiu et al. (2009) extendedthe study period. However, the time at which the glacier initiated accel-eration and the evolution of the frontal positions remain unknown.

Surface velocities near the terminus position could not be derived inthis study because of the low SNR in the offset tracking near the ice front.We consider that the faster ice motion near the terminus caused signifi-cant temporal changes in the surface features and hence their scatteringcharacteristics; the 46/35-day acquisition intervals in PALSAR/ASAR aretoo long for tracking the surface features of such a rapidly flowing regionwith sufficient correlation. However, the slower velocity data in theupstream could be derived; thus, we examined the temporal changesin velocities at the two sites closer to the terminus (A and B in Fig. 2a),in which the velocity data were acquired as frequently as possible. Weaveraged the velocities at 36 pixels (over 500 m2) for each site andassigned their standard deviations as the error bar in Fig. 2b. One sitewas in the main stream of Glaciar Upsala, and the other was in GlaciarBertachi, one of the western tributaries closest to the frontal position.Fig. 2b summarizes the temporal evolution of both surface velocities atthe two sites and frontal positions derived in this study. Fig. 3 showsthe terminus locations visually identified over time. Although we couldnot generate useful velocity data between 2006 and 2010 because of asatellite orbit problem, we observed a nearly 180% velocity increase be-tween 2005 and 2010 even at a point 7 km from the terminus in themain stream. From 2003 to 2005, the near-terminus velocities were2 m/day, which are in agreement with Skvarca et al.'s observations in2001 (Skvarca et al., 2003). Even in Glaciar Bertachi, it is evident thatthe 2010 surface velocities are nearly 200% those of the 2003–2005data. In 2010–2011, therefore, the near-terminus region is apparentlyflowing faster than previously recorded speeds. The frontal positionretreated nearly 3.5 km from 2002 to 2011 and approached the lowerend of Glaciar Bertachi. The speed up on Bertacchi between July 2010and January 2011 may be associated with the recent retreat. Althoughthe point A seems stable in Fig. 2b, it is likely that the similar speed-upat point A occurred but was not detected by the available data. In view

Fig. 2. (a) Spatial distribution of surface flow velocities at Glaciar Upsala, derived from PALSARfrontal position and velocities at sites A and B depicted in (a).

of the TerraSAR-X intensity image in Floricioiu et al. (2008), the frontalposition in January 2008 was close to that recorded in 2005. While thefront gradually retreated from 2002 to 2005 with possible seasonal fluc-tuations (Fig. 3), the retreating appears to have accelerated in August2008.

3.2. Glaciar Jorge Montt

Glaciar Jorge Montt is located at the northern end of the SPI andcalves into an unnamed fjord that opened as a result of its glacier retreatto a maximum of 19.5 km from 1898 to 2011 (Rivera, Koppes, Bravo, &Aravena, 2012). The maximum thinning rate was estimated tobe17.9 m/a between 1975 and 2000 (Rignot et al., 2003) and 21.5 ±0.8 m/a between 2000 and 2012 (Willis et al., 2012). In addition, therate of glacier terminus retreat was estimated to be 200 m/a between2001 and 2011 (Willis et al., 2012).

So far, few surface velocity measurements of Glaciar Jorge Montthave been reported. Rivera, Corripio, et al. (2012) derived an averagevelocity of 13 ± 4 m/day for the near-terminus region between 2010and 2011 by using both ground-based cameras and a feature trackingbased on ASTER images. Comparing Rignot et al.'s unpublished velocity

images obtained on January 4 and February 19, 2011. (b) Temporal changes in the average

Fig. 4. (a) Spatial distribution of surface flow velocities at Glaciar Jorge Montt, derived from PALSAR images obtained on January 11 and February 26, 2008. (b) Temporal changes in theaverage frontal position and velocities at profile A–B and site C in (a).

55M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

data derived from the 2004 C-band Radarsat-1 images, Rivera, Corripio,et al. (2012) suggested an acceleration between2004 and 2010. Howev-er, the surface velocity data in the other years and up-glacier partsremain uncertain.

Fig. 4a shows an example of a velocity map derived from dataacquired between January and February 2008. Although we could notderive the velocity data near the terminus region as was the case forGlaciar Upsala, the derived velocity maps complement the observationsof Rivera, Corripio, et al. (2012), because our velocity maps cover themiddle to upstream regions of the glaciers. Fig. 4b summarizes thetemporal evolutions of both frontal location changes and the meanvelocities at one flow velocity profile (A–B in Fig. 4a) and the otherupstream site (C in Fig. 4a). While the mean value of the A–B profile issmaller than the mean at the C, this is due to the averaging procedure,and the flow velocity at the center of A–B profile is higher than at C;the velocities and errors along the transverse profile are shown in theSupplementary material. The measured velocity profile is located

Fig. 5. Temporal evolution of the frontal position at Glaciar JorgeMontt between 2002 and2011. Background is a PALSAR-based scattering intensity image acquired on February 19,2011.

5.5 km upstream from that measured by Rivera, Corripio, et al. (2012).Fig. 5 indicates the frontal locations identified at each epoch. The rela-tionship between the retreating rate and flow velocities from 2003 to2007 remains uncertain because not all of the image pairs allowed usto derive the flow velocity data. However, both the retreating rate andflow velocities are not constant over time. Although the frontal positioncontinued to retreat since 2003, it became stagnant from 2007 to 2008when the flow velocities significantly decreased. From 2008 onward,the retreating rate apparently accelerated, indicating ~470 m/a, andthe flow velocities in 2011 increased by 130%, compared to thoserecorded in 2003. While the flow acceleration reported by Rivera,Corripio, et al. (2012) was derived by comparing the data acquired in2004 and 2010, Fig. 4b suggests that the acceleration was initiated in2008.

3.3. Glaciar Occidental

Glaciar Occidental is located on thewestern side of the SPI and flowsinto a proglacial lake. On the basis of the analysis of Landsat TM andAdvanced Visible and Near Infrared Radiometer-2 (AVNIR-2) on ALOSimages, the terminus retreat between 1986 and 2006 was estimatedto be 1.1 km, and the terminus shape was found to be disintegrated(JAXA, 2010). Regarding surface elevation change rate, Glaciar Occiden-tal is also thinning at a rate of 2–5 m/a (Willis et al., 2012). To ourknowledge, however, no flow velocity data have been published so far.

We selected two transverse profiles at downstream and upstreampositions (A–B and C–D, respectively, in Fig. 6a). In contrast to therapidly flowing glaciers previously described, this glacier allowedus to derive flow velocity data even near the terminus, likely becausethe slower flow velocities allowed the glacier to maintain its surfacefeatures over time. Fig. 6b summarizes the temporal evolution ofmean velocities at each profile and the frontal position; temporalevolution of the velocity profile is shown in the Supplementarymaterial. While the velocities in the upstream were stable overtime (C–D), those near the terminus in 2008–2010 have apparentlyincreased by more than 200% over 2004–2005. The near-terminusregion apparently underwent a significant longitudinal stretching.We evaluated the frontal position retreat to be ~2 km from 2004 to2011. However, the frontal positions at some epochs are missingbecause a vague image contrast prevented clear identification. In-stead of a clear frontal position, we observed numerous icebergsthat likely disintegrated from the terminus. The rate of terminus

Fig. 6. (a) Spatial distribution of surface flow velocities at Glaciar Occidental, derived from PALSAR images obtained on January 4 and February 19, 2011. (b) Temporal changes in theaverage frontal position and velocities at profile A–B and site C–D in (a).

56 M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

retreat accelerated from 55 m/a for the period from 1986 to 2006 to~285 m/a for the recent period from 2004 to 2011.

3.4. Glacier Pio XI

Glaciar Pio XI, the largest glacier in the SPI, flows to the west from alocation near Volcán Lautaro (elevation 3380 m) and bifurcates intotwo tongues to the north and south near its lower end, which flow

Fig. 7. (a) Spatial distribution of surface flow velocities at Glaciar Pio XI, derived from PALSARvelocity at each of the three profiles (A–B, C–D, and E–F) is shown in (b–d). (b) Temporal chan(c) Temporal changes in the average frontal position and velocity across profile C–D in the souupstream.

into ‘Lago Greve’ and Eyre Fjord, respectively (Fig. 7a); the southernfront is thus a tidewater front. In contrast to the other glaciers in Patago-nia that are rapidly retreating, Glaciar Pio XI was known for its signifi-cant terminus advances at both tongues during the 20th century(Rivera, Lange, Aravena, & Casassa, 1997; Warren & Rivera, 1994). Thefront region is reportedly thickening at a rate of greater than 2 m/a(Rivera & Casassa, 1999; Willis et al., 2012). On the basis of ground-based measurements recorded in November 1995, Rivera et al. (1997)

images obtained on January 4 and February 19, 2011. Temporal evolution in the averageges in the average frontal position and velocity across profile A–B in the northern tongue.thern tongue. (d) Temporal changes in the average velocity across profile E–F in the main

57M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

reported maximum velocities of 17 to 50 m/day near the southern ter-minus. Although these faster velocities, in addition to the terminusadvances, have led to an interpretation of a surging glacier, the factorscontrolling the dynamics of Glaciar Pio XI remain uncertain (Riveraet al., 1997). Apparently, this glacier has not been directly respondingto the climate change (Warren & Rivera, 1994).

We selected three profiles across the northern tongue (A–B inFig. 7a), southern tongue (C–D in Fig. 7a), and the main upstreamregion (E–F in Fig. 7a) to reveal the temporal evolution of the sur-face velocities. Fig. 7b–d summarizes the temporal evolution of themean surface velocities at each profile, and the frontal positionchanges at the northern and southern tongues from 2003 to 2011are shown in Fig. 7b and c. The temporal evolution of each velocityprofile is shown in the Supplementary material. The flow velocityin the main upstream was relatively stable over time, although itrevealed deceleration in 2007 (Fig. 7d). In contrast, the velocitiesat the northern and southern tongues were highly variable overtime (Fig. 7b and c). The southern tongue reveals faster signals in2003, 2005, and 2007, when the velocities nearly doubled those inslower phases. While the northern tongue was slowly flowing at0.2 m/day in 2003, it gradually accelerated and reached 2.0 m/dayin 2007. The upstream thus does not appear to directly control theflow velocities in the downstream.

Despite the highly variable velocity changes to the northern andsouthern tongues (Fig. 7b and c), we did not observe any significantretreat as observed in Glaciars Upsala, Jorge Montt, and Occidental.Fig. 8 includes the frontal positions over the analyzed period. Whilethese positions have overall advanced from 2003 to 2011, we observedfluctuations in the front positions within a range of ~250 m betweenadvancing and retreating periods.

4. Discussion and conclusion

Although the retreat and thinning of glaciers in the SPI over the pastdecades have been well-documented, we used SAR image analysis todemonstrate that not all of the examined glaciers revealed significantchanges in frontal positions and flow velocities from 2002 to 2011.Moreover, even for such glaciers that underwent rapid retreat andacceleration, the terminus retreat and flow acceleration occurredsporadically rather than gradually over time, suggesting that positive-feedback processes promoted such movement. This finding was madepossible through the use of SAR images that were free from cloud andnight time problems, although more frequent image acquisitions aredesirable.

Fig. 8. Temporal evolution of the frontal position at Glaciar Pio XI between 2002 and 2011.Background is a PALSAR-based scattering intensity image acquired on February 19, 2011.

The observed rapid retreat and acceleration at Glaciars Upsala, JorgeMontt, and Occidental have important implications for calving mecha-nisms. In particular, despite the marked differences in absolute flowvelocity among the three glaciers, the region toward the terminus ateach glacier underwent significant acceleration, which was observedby calculating differences in the velocity map at two epochs (Fig. 9).Fig. 9 indicates larger velocity increments downstream, which demon-strates longitudinal stretching or extension toward the termini.

Because longitudinal stretching is observed even at distances greaterthan ~5 km upstream from the 2011 termini, the actual accelerationsnear the terminus regions of Glaciars Upsala and Jorge Montt will prob-ably be much larger (Fig. 9a and b). The near-terminus area in GlaciarOccidental flowed at a rate of ~0.5 m/day by at least 2005 and beganto accelerate in 2007, while the upstream velocities were stable. Suchlongitudinal stretching toward the terminus is clearly revealed atGlaciarOccidental (Fig. 9c).

The observed longitudinal stretching at the rapidly retreating gla-ciers will contribute to increasing the crevasse-depths, and thus appearsto support a calving model based on crevasse-depth criteria (Benn,Hulton, et al., 2007; Benn, Warren, et al., 2007; Nick et al., 2010), inwhich the absolute flow velocity itself is not critical. Benn, Hulton,et al. (2007) proposed that longitudinal stretching due to the spatialvelocity gradient, which determines the location and depth of surfacecrevasses, and the difference between ice thickness and water depth(effective pressure) are the primary controls for the frontal position ofcalving. This theory assumes that the basal sliding velocity is equal tothe vertically averaged ice velocity. While Benn et al.'s model predictsthe calving front position in which the depth of the surface crevasseequals the ice height above the water line, Nick et al. (2010) modifiedthe model to predict calving for cases in which both surface and basalcrevasses penetrate the full thickness of the glacier.

The longitudinal stretching and rapid terminus retreat at GlaciarUpsala, Jorge Montt, and Occidental can be regarded as indicators ofthe on-going dynamic thinning process. These three glaciers havebeen thinning over the past decade (Willis et al., 2012). Because theglacier thinning reduces the viscous ice velocity due to the decrease ofgravitational force, we attribute the observed speed-up to the enhancedbasal sliding velocity.We consider that the basal sliding enhancement iscaused by the surface melt input and subsequent reduction of effectivepressure due to the water pressure increase as observed at Greenland(e.g., Sundal et al., 2011). If the basal sliding velocity dominantly contrib-utes to the vertically averaged velocity, the longitudinal stretching willalso increase and allow the crevasses to penetrate further, causing theterminus to retreat up-glacier. As recognized from a bathymetric surveyconducted at Glaciar Upsala (Naruse & Skvarca, 2000), this terminusretreatwill likely be halted in shallowerwater because the effective pres-sure will increase. Such stabilization of the terminus retreat is, however,not simply controlled by the local water-depth or ice thickness as in theheight-above-buoyancy model by van der Veen (1996). Physics-basedcrevasse-depth calving criteria can eliminate the unrealistically largeterminus retreats predicted from the height-above-buoyancy criterion(Nick et al., 2010).

Our observations suggest that even other seemingly stable glaciersbetween 2003 and 2011 may undergo rapid retreat and accelerationin the future. The Glaciar Occidental observation provides an importantlesson for understanding the sporadic nature of calving processes. How-ever, we cannot determine the point at which glaciers will undergosignificant frontal retreat and acceleration in the future because calvingprocesses involve a variety of unknown variables, which remain poorlyunderstood (Benn, Warren, et al., 2007). Continuation of frequent sur-face velocity monitoring and combining the results with bathymetricand ice-thickness survey data can more precisely validate and refinethe calving model, which may enable prediction of the initiation ofdynamic-thinning.

The behavior of Glaciar Pio XI is enigmatic, and has been suggestedas a surging type (Rivera et al., 1997). Transverse velocity profiles

Fig. 9. Velocity increases observed at (a) Glaciar Upsala from January 2003 to January 2011, (b) Glaciar Jorge Montt from January 2003 to January 2011, and (c) Glaciar Occidental fromJanuary 2003 to January 2011.

58 M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

derived in this study indicate parabolic shapes with smaller velocitiesnear the edge, which increase toward the center and do not exhibitthe plug flow observed in other surging glaciers (Kamb et al., 1985;Murray, Strozzi, Luckman, Jiskoot, & Christakos, 2003; Yasuda &Furuya, 2013). However, note that the maximum water depths infront of the calving front at Eyre Fjord are ~20 m, and that those atLago Greve are speculated to be much shallower than 150 m (Warren& Rivera, 1994). These water depths are significantly shallower thanthose at other glaciers, which may be attributed to the former surgingepisodes that should have transported a sufficient amount of sedimentsto raise the bottom depth of the fjord. While it appears unlikely that arapid terminus retreat will begin in the near future, it is important tocontinue monitoring Glaciar Pio XI with the same frequency as that ofother calving glaciers in the SPI.

It remains uncertain how much the total ice loss in the PatagonianIce Fields is split into surface processes (runoff and precipitation) andice dynamics, and the surface velocity data are indispensable to evaluatethe contribution from ice dynamics. The velocities presented in thisstudy are still not complete, because the velocities at the very front ofthe glaciers are missing. Because the missing velocities near the frontare probably faster, the inferred ice velocities might be helpful to con-strain the lower bound of ice discharge. Nonetheless, given the factthat not all the examined glaciers revealed acceleration and terminusretreat, the contribution from ice dynamics to the total ice loss mightbe not as high as those in Greenland, wherewidespread ice accelerationwas observed (Joughin et al., 2008) and the partitioning ratio wasshown to be equal (van den Broeke et al., 2009). In order to quantifythe ice discharge, both surface mass balance and ice thickness data inthe Patagonian Ice Fields are also necessary.

Acknowledgments

PALSAR level 1.0 data in this study were provided by the PALSAR In-terferometry Consortium to Study our Evolving Land surface (PIXEL) andthe ALOS 3rd PI project under cooperative research contracts with JAXA.The PALSAR data are owned by JAXA and theMinistry of Economy, Tradeand Industry. Envisat data are copyrighted by ESA and were providedunder Cat-1 project 9364. This study is partially supported by KAKENHI

(24651001).We acknowledge two anonymous reviewers for their help-ful and constructive comments.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.rse.2013.07.034.

References

Andersen, P. H., Aksnes, K., & Skonnord, H. (1998). Precise ERS-2 orbit determinationusing SLR, PRARE, and RA observations. Journal of Geodesy, 72(7/8), 421–429.

Aniya, M., Sato, H., Naruse, R., Skvarca, P., & Casassa, G. (1996). The use of satellite andairborne imagery to inventory outlet glaciers of the Southern Patagonia Icefield,South America. Photogrammetric Engineering and Remote Sensing, 62(12), 1361–1369.

Aniya, M., Sato, H., Naruse, R., Skvarca, P., & Casassa, G. (1997). Recent glacier variations inthe Southern Patagonia Icefield, South America. Arctic and Alpine Research, 29, 1–12.

Aniya, M., & Skvarca, P. (1992). Characteristics and variations of Upsala and Morenoglaciers, southern Patagonia. Bulletin of Glacier Research, 10, 39–53.

Benn, D. I., Hulton, N. R. J., & Mottram, R. H. (2007). ‘Calving laws’, ’sliding laws’ and thestability of tidewater glaciers. Annals of Glaciology, 46, 123–130.

Benn,D. I.,Warren, C. R., &Mottram, R.H. (2007). Calving processes and thedynamics of calv-ing glaciers. Earth-Science Reviews, 82, 143–179, http://dx.doi.org/10.1016/j.earscirev.2007.02.002.

Bevan, S. L., Murray, T., Luckman, A. J., Hanna, E., & Huybrechts, P. (2012). Stable dynamicsin a Greenland tidewater glacier over 26 years despite reported thinning. Annals ofGlaciology, 53(60), 241–248, http://dx.doi.org/10.3189//2102AoG60A076.

Chen, J. L., Wilson, C. R., Tapley, B.D., Blankenship, D.D., & Ivins, E. R. (2007). Patagonia icefield melting observed by Gravity Recovery and Climate Experiment (GRACE).Geophysical Research Letters, 34, L22501, http://dx.doi.org/10.1029/2007GL031871.

Ciappa, A., Pietranera, L., & Battazza, F. (2010). Perito Moreno Glacier (Argentina) flowestimation by COSMO SkyMed sequence of high-resolution SAR-X imagery. RemoteSensing of Environment, 114, 2088–2096.

Cuffey, K. M., & Paterson, W. S. B. (2010). The physics of glaciers (4th ed.): Elsevier.Davies, B. (2012). GLIMS Glacier Database. Boulder, CO: National Snow and Ice Data

Center/World Data Center for Glaciology. Digital Media.Delgado, S. (2009). GLIMS glacier database. Boulder, CO: National Snow and Ice Data

Center/World Data Center for Glaciology. Digital Media.Dickmann, N. (2007). GLIMS glacier database. Boulder, CO: National Snow and Ice Data

Center/World Data Center for Glaciology. Digital Media.Floricioiu, D., Eineder, M., Rott, H., & Nagler, T. (2008). Velocities of major outlet

glaciers of the Patagonia Icefield observed by TerraSAR-X. Proc. IGARSS 2008(pp. IV-347–IV-350), http://dx.doi.org/10.1109/IGARSS.2008.4779729.

Floricioiu, D., Eineder, M., Rott, H., Yague-Martinez, N., & Nagler, T. (2009). Surface velocityand variations of outlet glaciers of the Patagonia Icefields by means of TerraSAR-X.Proc. IGARSS 2009 (pp. II-1028–II-1031), http://dx.doi.org/10.1109/IGARSS.2009.5418279.

59M. Muto, M. Furuya / Remote Sensing of Environment 139 (2013) 50–59

Goldstein, R. M., Engelhardt, H., Kamb, B., & Frolich, R. M. (1993). Satellite radar inter-ferometry for monitoring ice sheet motion: Application to an Antarctic icestream. Science, 262(5139), 1525–1530.

Gourmelen, N., Kim, S. W., Shepherd, A., Park, J. W., Sundal, A. V., Björnsson, H., et al. (2011).Ice velocity determined using conventional and multiple-aperture InSAR. Earth andPlanetary Science Letters, 307, 156–160, http://dx.doi.org/10.1016/j.epsl.2011.04.026.

Holland, D.M., Thomas, R. H., de Young, B., Ribergaard, M. H., & Lyberth, B. (2008).Acceleration of Jakobshavn Isbr triggered by warm subsurface ocean waters.Nature Geoscience, 1, 659–664, http://dx.doi.org/10.1038/ngeo316.

Ivins, E. R.,Watkins, M. M., Yuan, D. N., Dietrich, R., Casassa, G., & Rülke, A. (2011). On-landice loss and glacial isostatic adjustment at the Drake Passage; 2003–2009. Journal ofGeophysical Research, 116, B02403, http://dx.doi.org/10.1029/2010JB007607.

Jacob, T., Wahr, J., Pfeffer, W. T., & Swenson, S. (2012). Recent contributions of glaciers andice caps to sea level rise.Nature, 482, 514–518, http://dx.doi.org/10.1038/nature10847.

Jarvis, A., Reuter, H. I., Nelson, A., & Guevara, E. (2008). Hole-filled seamless SRTM data V4,International Centre for Tropical Agriculture (CIAT). available from. http://srtm.csi.cgiar.org

JAXA (2010). Significant retreats of huge glaciers in Patagonia, South America (Part 3).available at. http://www.eorc.jaxa.jp/en/imgdata/topics/2010/tp100421.html

Joughin, I., Das, S. B., King, M.A., Smith, D. E., Howat, I. M., & Moon, T. (2008). Seasonalspeedup along the Western Flank of the Greenland ice sheet. Science, 320, 781–783,http://dx.doi.org/10.1126/science.1153288.

Joughin, I., Kwok, R., & Fahnestock, M. (1996). Estimation of ice-sheet motion usingsatellite radar interferometry: Method and error analysis with application toHumboldt Glacier, Greenland. Journal of Glaciology, 42, 564–575.

Kamb, B., Raymond, C. F., Harrison, W. D., Engelhardt, H., Echelmeyer, K. A.,Humphrey, N., et al. (1985). Glacier surge mechanism: 1982–1983 surge ofVariegated Glacier, Alaska. Science, 227, 469–479.

Katagiri, S., & Yamamoto, Y. (2008). Technology of precise orbit determination. FujitsuScientific & Technical Journal, 44(4), 401–409.

Kobayashi, T., Takada, T., Furuya, M., & Murakami, M. (2009). Locations and types of rup-tures involved in the 2008 Sichuan earthquake inferred from SAR image matching.Geophysical Research Letters, 36, L07302, http://dx.doi.org/10.1029/2008GL036907.

Masiokas, M. (2009). GLIMS glacier database. Boulder, CO: National Snow and Ice DataCenter/World Data Center for Glaciology. Digital Media.

Masiokas, M. (2010). GLIMS glacier database. Boulder, CO: National Snow and Ice DataCenter/World Data Center for Glaciology. Digital Media.

Michel, R., Avouac, J. -P., & Taboury, J. (1999). Measuring ground displacementsfrom SAR amplitude images: Application to the Landers Earthquake. GeophysicalResearch Letters, 26, 875–878.

Michel, R., & Rignot, E. (1999). Flow of Glacier Moreno, Argentina, from repeat-passShuttle Radar images: Comparison of the phase correlation method with radarinterferometry. Journal of Glaciology, 45(149), 93–100.

Moon, T., & Joughin, I. (2008). Changes in ice front position on Greenland's outlet glaciersfrom 1992 to 2007. Journal of Geophysical Research, 113, F02022, http://dx.doi.org/10.1029/2007JF000927.

Moon, T., Joughin, I., Smith, B., & Howat, I. (2012). 21st-century evolution of Greenlandoutlet glacier velocities. Science, 336, 576–578, http://dx.doi.org/10.1126/science.1219985.

Murray, T., Strozzi, T., Luckman, A., Jiskoot, H., & Christakos, P. (2003). Is there a single surgemechanism? Contrasts in dynamics between glacier surges in Svalbard andother regions. Journal of Geophysical Research, 108, B52237, http://dx.doi.org/10.1029/2002JB001906.

Naruse, R., & Skvarca, P. (2000). Dynamic features of thinning and retreating GlaciarUpsala, a lacustrine calving glacier in Southern Patagonia. Arctic, Antarctic, and AlpineResearch, 32(4), 485–491.

Naruse, R., Skvarca, P., Kadota, T., & Koizumi, K. (1992). Flow of Upsala and MorenoGlaciers, southern Patagonia. Bulletin of Glacier Research, 10, 55–62.

Naruse, R., Skvarca, P., Satow, K., Takeuchi, Y., & Nishida, K. (1995). Thickness change andshort-term flow variation of Moreno Glacier, Patagonia. Bulletin of Glacier Research,13, 21–28.

Naruse, R., Skvarca, P., & Takeuchi, Y. (1997). Thinning and retreat of Glaciar Upsala, andan estimate of annual ablation changes in southern Patagonia. Annals of Glaciology,24, 38–42.

Nick, F. M., van der Veen, C. J., Vieli, A., & Benn, D. I. (2010). A physically based calvingmodel applied to marine outlet glaciers and implications for the glacier dynamics.Journal of Glaciology, 56, 781–794, http://dx.doi.org/10.3189/002214310794457344.

Pritchard, H. D., Arthern, R. J., Vaughan, D.G., & Edwards, L. A. (2009). Extensive dynamicthinning on the margins of the Greenland and Antarctic ice sheets. Nature, 461,971–975, http://dx.doi.org/10.1038/nature08471.

Rignot, E. (2008). Changes in West Antarctic ice stream dynamics observed with ALOSPALSAR data. Geophysical Research Letters, 35, L12505, http://dx.doi.org/10.1029/2008GL033365.

Rignot, E., & Kanagaratnam, P. (2006). Changes in the velocity structure of the Greenlandice sheet. Science, 311, 986–990, http://dx.doi.org/10.1126/science.1121381.

Rignot, E., Koppes, M., & Velicogna, I. (2010). Rapid submarine melting of the calving facesof West Greenland glaciers. Nature Geoscience, 3, 187–191, http://dx.doi.org/10.1038/ngeo765.

Rignot, E., Rivera, A., & Casassa, G. (2003). Contribution of the Patagonia Icefields of SouthAmerica to sea level rise. Science, 302(5644), 434–437.

Rivera, A., & Casassa, G. (1999). Volume changes on Pio XI glacier, Patagonia: 1975–1995,Global Planet. Change, 22, 233–244.

Rivera, A., Corripio, J., Bravo, C., & Cisternas, S. (2012). Glaciar Jorge Montt (ChileanPatagonia) dynamics derived from photos obtained by fixed cameras and satelliteimage feature tracking. Annals of Glaciology, 53(60), 147–155, http://dx.doi.org/10.3189/2012AoG60A152.

Rivera, A., Koppes, M., Bravo, C., & Aravena, J. C. (2012). Little ice age advance and retreatof Glaciar Jorge Montt, Chilean Patagonia. Climate of the Past, 8, 403–414, http://dx.doi.org/10.5194/cp-8-403-2012.

Rivera, A., Lange, H., Aravena, J. C., & Casassa, G. (1997). The 20th-century advance ofGlacier Pio XI, Chilean Patagonia. Annals of Glaciology, 24, 66–71.

Rott, H., Stuefer, M., Siegel, A., Skvarca, P., & Eckstaller, A. (1998). Mass fluxes anddynamics of Moreno Glacier, Southern Patagonia Icefield. Geophysical ResearchLetters, 25(9), 1407–1410.

Sakakibara, D., Sugiyama, S., Sawagaki, T., Marinsek, S., & Skvarca, P. (2013). Rapid retreat,acceleration and thinning of Glaciar Upsala, Southern Patagonia Icefield, initiated in2008.Annals of Glaciology, 54(63), 131–138, http://dx.doi.org/10.3189/2013AoG63A236.

Sansosti, E., Berardino, P., Manunta, M., Serafino, F., & Fornaro, G. (2006). Geometrical SARimage registration. IEEE Transactions on Geoscience and Remote Sensing, 44(10),2861–2870.

Scharroo, R., & Visser, P. (1998). Precise orbit determination and gravity field improve-ment for the ERS satellites. Journal of Geophysical Research, 103(C4), 8113–8128.

Skvarca, P., Raup, B., & De Angelis, H. (2002). Calving rates in fresh water: New data fromsouthern Patagonia. Annals of Glaciology, 34, 379–384.

Skvarca, P., Raup, B., & De Angelis, H. (2003). Recent behaviour of Glacier Upsala, afast-flowing calving glacier in LagoArgentino, southernPatagonia. Annals of Glaciology,36, 184–188.

Skvarca, P., Satow, K., Naruse, R., & Leiva, J. C. (1995). Recent thinning, retreat and flow ofUpsala Glacier, Patagonia. Bulletin of Glacier Research, 13, 11–20.

Straneo, F., Curry, R. G., Sutherland, D. A., Hamilton, G. S., Cenedese, C., Vge, K., et al.(2011). Impact of fjord dynamics and glacial runoff on the circulation near HelheimGlacier. Nature Geoscience, 4, 322–327, http://dx.doi.org/10.1038/ngeo1109.

Strozzi, T., Kouraev, A., Wiesmann, A., Wegmüller, U., Sharov, A., & Werner, C. L. (2008).Estimation of Arctic glacier motion with satellite L-band SAR data. Remote Sensingof Environment, 112, 636–645.

Strozzi, T., Luckman, A., Murray, T., Wegmüller, U., & Werner, C. (2002). Glacier motionestimation using SAR offset-tracking procedures. IEEE Transactions on Geoscienceand Remote Sensing, 40(11), 2384–2391.

Stueffer, M., Rott, H., & Skvarca, P. (2007). Glaciar Perito Moreno, Patagonia: Climate sen-sitivities and glacier characteristics preceding the 2003/04 and 2005/06 dammingevents. Journal of Glaciology, 53(180), 3–16.

Sugiyama, S., Skvarca, P., Naito, N., Enomoto, H., Tsutaki, S., Tone, K., et al. (2011). Icespeed of a calving glacier modulated by small fluctuations in basal water pressure.Nature Geoscience, 4, 597–600, http://dx.doi.org/10.1038/ngeo1218.

Sundal, A. V., et al. (2011). Melt-induced speed-up of Greenland ice sheet offset by efficientsubglacial drainage. Nature, 469, 521–524, http://dx.doi.org/10.1038/nature09740.

van den Broeke, M., Bamber, J., Ettema, J., Rignot, R., Schrama, E., van de Berg, W. J.,et al. (2009). Partitioning recent Greenland mass loss. Science, 326, 984–986,http://dx.doi.org/10.1126/science.1178176.

van der Veen, C. J. (1996). Tidewater calving. Journal of Glaciology, 42(141), 375–385.Warren, C., & Aniya, M. (1999). The calving glaciers of southern South America. Global and

Planetary Change, 22, 59–77.Warren, C., & Rivera, A. (1994). Non-linear climatic response of calving glaciers: A case study

of Pio XI Glacier, Chilean Patagonia. Revista Chilena de Historia Natural, 67, 385–394.Willis, M. J., Melkonian, A. K., Pritchard, M. E., & Rivera, A. (2012). Ice loss from the Southern

Patagonian Ice Field, South America, between 2000 and 2012. Geophysical Research Let-ters, 39, L17501, http://dx.doi.org/10.1029/2012GL053136.

Yasuda, T., & Furuya, M. (2013). Short-term glacier velocity changes atWest Kunlun Shan,Northwest Tibet, detected by Synthetic Aperture Radar data. Remote Sensing of Envi-ronment, 128, 87–106, http://dx.doi.org/10.1016/j.rse.2012.09.021.