InSAR-based Glacier Velocity Mapping in the Parlung Zangbo …sar.kangwon.ac.kr/paper/kjrs_kecq.pdf · 2019. 2. 28. · abundant rainfall at ~1000-3000 mm above the glacier mass equilibrium

1. Introduction

Mountain glaciers are highly sensitive to climatechange and serve as an important indicator for climatechange (McCarthy et al., 2011). From a globalperspective, the majority of glaciers are in a state ofretreat, and this phenomenon is particularly evident in

mountain glaciers (Barry, 2006). Although a smallnumber of glaciers have advanced over the past 50years, 82% of the more than 5000 glaciers in WestChina have experienced a state of retreat, resulting in a4.5% reduction in total glacier area (Jiang et al., 2011).On the Tibetan Plateau, the majority of monitoredglaciers are showing a trend of retreat (Yao et al., 2012).

InSAR-based Glacier Velocity Mapping in the Parlung Zangbo River Basin,

Tibetan Plateau, China

Chang-Qing Ke 1)†·Hoonyol Lee 2)·Lan-Yu Li3)

Abstract: By applying the method of SAR interferometry to X-band synthetic aperture radar (SAR) image ofCOSMO-SkyMed, detailed motion patterns of five glaciers in the Parlung Zangbo River basin, Tibetan Plateau, inJanuary 2010 have been derived. The results indicate that flow patterns are generally constrained by the valleygeometry and terrain complexity. The maximum of 123.9 ma-1 is observed on glacier No.1 and the minimum of39.4 ma-1 is found on glacier No.3. The mean values of five glaciers are between 22.9 and 98.2 ma-1. Glaciers No.1,No.2, No.4 and No.5 exhibit high velocities in their upper sections with big slope and low velocities in the lowersections. A moraine lake accelerates the speed of mass exchange leading to a fast flow at the terminal of glacierNo.3. These glaciers generally move along the direction of decreased elevation and present a macroscopic illustrationof the motion from the northwest to the southeast. The accuracy of DEM and registration conditions of DEM-simulated terrain phases has certain effects on calculations of glacier flow direction and velocity. The error field isrelatively fragmented in areas inconsistent with the main flow line of the glaciers, and the shape and uniformity ofglacier are directly related to the continuous distribution of flow velocity errors.

Key Words: InSAR method, Glacier velocity, Error analysis, Parlung Zangbo River basin

Korean Journal of Remote Sensing, Vol.35, No.1, 2019, pp.15~28https://doi.org/10.7780/kjrs.2019.35.1.2

ISSN 1225-6161 ( Print )ISSN 2287-9307 (Online)

Article

Received September 28, 2018; Revised October 15, 2018; Accepted October 30, 2018; Published online November 22, 20181) Professor, School of Geography and Ocean Science, Nanjing University2) Professor, Department of Geophysics, Kangwon National University3) Postdoctor, School of Geography and Ocean Science, Nanjing University

†Corresponding Author: Chang-Qing Ke ([email protected])This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License(http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction inany medium, provided the original work is properly cited.

– 15 –

02Chang-Qing Ke(15~28)(All컬러)ok.qxp_원격35-1(2019) 2019. 2. 27. 오후 2:02 페이지 15

https://orcid.org/0000-0003-0212-4069

http://orcid.org/0000-0002-5980-8640

Rises in temperature and rainfall precipitation (frequencyof rainfall is higher than snow at high latitudes and lowelevations) are the primary factors causing the meltingof glaciers (Huai et al., 2014), and they have also ledto a negative correlation between glacier mass balanceand time scale (Kang et al., 2010).

In the context of global warming, real-timemonitoring of physical changes and motion patterns ofglaciers is necessary for predicting glacier evolutionand future climate (Diolaiuti et al., 2012). Glaciersurface flow velocity is an important physical factor formass balance changes and a critical parameter forglacier dynamics research. Accurate measurements of seasonal and inter-annual glacier velocity fields areof great importance to understanding the motiondisplacement of glaciers and ice sheets caused byclimate change (Berthier et al., 2005). In addition,glacier motion provides important information foridentifying the glacier mass balance equilibrium lineand sea level rise (Dowdeswell et al., 1999).

Field measurements have long been an importantmethod of generating surface flow velocity field mapsof glaciers. Compared with traditional measurementsusing a measuring stake, acquiring surface flowvelocity fields by remote sensing methods whichgenerate data on a large scale, is more convenient andhas a lower cost (Yan et al., 2013). The syntheticaperture radar (SAR) possesses all-weather operationalcapabilities of obtaining images in glacier regions andis not affected by clouds or rains, and it has rapidlybecome a frontier area in remote sensing applications(Cigna et al., 2012). Because glacier motion in SARimage can show distinct variation characteristics,numerous studies have demonstrated the applicationadvantages of SAR data and SAR interferometry(InSAR) techniques in the field of glacier dynamics (Liet al., 2008). Satellites carrying SAR sensors have beenlaunched in a number of countries, and the increasinglynumerous SAR data provide more additionalinformation for applied research in glaciology (Jiang et

al., 2011). As an efficient and sensitive extractiontechnology of surface flow velocity field, the wide useof InSAR methods provides rapid and accuratereference information for studies of glacier dynamics(Michel and Rignot, 1999). The majority of previousstudies have focused on motion field evaluations ofpolar glaciers and ice sheet mapping, such as in Alaska,Greenland and Antarctica (Ciappa et al., 2010;Goldstein et al., 1993; Joughin, 2002; Mohr et al.,1998). In addition, the application of InSAR techniqueshas obtained excellent result for glaciers in the southregions of the Himalayas, Karakoram, TianshanMountains and Yigong Zangbo River basin (Jiang etal., 2012; Ke et al., 2013; Kumar et al., 2011; Yan et al.,2013). Glacier observations have potential implicationsfor water resource utilization, local climate research andnatural disaster predictions (Liu et al., 2012). Since2006, China has been conducting comprehensiveobservational studies of glacier changes in the ParlungZangbo River basin in southeast Tibet (Yang et al.,2010). However, research on glacial motion is still rare.

Compared with C-band and L-band images, InSARanalysis of X-band images on short time scales canachieve more accurate image pair matching and terrainregistration, thus improving the quality of InSARresults (Ciappa et al., 2010; Crosetto et al., 2015;Schubert et al., 2013) . Therefore, the use of X-bandimages has a great advantage in defining glacierboundary and extracting surface flow velocity fields.In this study, the velocity fields of five glaciers in theParlung Zangbo River basin are mapped withCOSMO-SkyMed (Constellation of small Satellites forMediterranean basin Observation) X-band data. Theglacier motion patterns, height and position of theirmass balance equilibrium lines are described by usingthe InSAR technique and accumulation area ratio(AAR) method. Combined with the physical parametersof the COSMO-SkyMed satellite observation system,the accuracies of glacier velocity fields are evaluated,and distribution field of glacier velocity errors are

Korean Journal of Remote Sensing, Vol.35, No.1, 2019

– 16 –


mapped. The results provide a reference for researchon glacier change and rational use of glacier waterresources in the southeast Tibetan Plateau.

2. Study area and data

1) Study areaThe Parlung Zangbo River basin lies in the southeast

Tibetan Plateau at 30.16°-30.40°N, 94.50°-95.20°E(Fig. 1). The basin is located at the junction where theHimalayas traverse along the Tanggula and HengduanMountains, and it covers a vast area, with theNamchabarwa Peak as the centre and Great Bend of Brahmaputra lower reaches as the arc. Thegeomorphological features of the basin area aregenerally high mountains, deep valleys and widegullies in an alternative distribution, and the terrain isgreatly undulating from 5500 m to 2479 m above sealevel (Zeng et al., 2007). The Parlung Zangbo Riverbasin belongs to the subtropical mountainous climate,and it is influenced by the southwest monsoon, withabundant rainfall at ~1000-3000 mm above the glaciermass equilibrium line and mean temperature in summerabove 1°C (Chen et al., 2012).

The Parlung Zangbo River is the largest tributary tothe north of Brahmaputra’s Great Bend, and itoriginates from the Laigu glacier (also known asYanong glacier) and flows from the east to the westthrough Ranwu, Songzong and Guxiang. At Tongmai,the Parlung Zangbo River converges with the YigongZangbo River from the west and turns to the southwest,entering the Brahmaputra River at ~40 km. The ParlungZangbo River is the largest distribution area ofmaritime glacier in China, and it retains a large rangeof glacial remains. A total of 2968 modern glaciers arein the Parlung Zangbo River basin, with an averagearea and thickness of 2.14 km2 and 109 m, respectively.The majority of these glaciers are in a state of pressuremelting because of intense surface melting and fastmovement (Li et al., 1986). The five glaciers in thestudy area are designated Glaciers No.1-No.5 (Fig. 1).

2) DataThe COSMO-SkyMed satellite observation system

is a radar satellite constellation jointly developed by theItalian Space Agency and Defence Department. Itconsists of four medium-sized X-band (λ = 3.1 cm)near-Earth satellites at a temporal resolution of 16 days,and it is completely launched by the end of 2010(COSMO SkyMed System, 2010). COSMO-SkyMedimages are characterised by high spatial resolution,wide coverage range, selectable incidence angles, andmulti-polarization. Combined with precise orbit data,COSMO-SkyMed images have a great advantage inmonitoring ground surface deformations using repeat-track interferometry. In local areas, X-band images maybe more sensitive and can identify tiny surfacedeformations (Ciappa et al., 2010; Liu et al., 2011). Weinvestigate the characteristics of glacier surface flowvelocity using two single-look-complex (SLC) SARimages (Table 1) in January 2010 on STRIPMAPHIMAGE mode that covers the Parlung Zangbo Riverbasin at a spatial resolution of 3 m. Because of the timeinterval of only 1 day, the interference fringes are

InSAR-based Glacier Velocity Mapping in the Parlung Zangbo River Basin, Tibetan Plateau, China

– 17 –

Fig. 1. The location of study area and five glaciers, overlyingX-band COSMO-SkyMed SAR image.


relatively clear, even in the southeast Tibetan Plateau,which has undulating terrain and complex mountainstructures.

Shuttle Radar Topography Mission (SRTM) digitalelevation model (DEM) at a spatial resolution of 90 ×90 m (version 4) is used for topographic corrections,image registration and geocoding of multi-look SARimages. Additionally, the DEM is adjunctively usedto identify the glacier boundary, calculate the slopeand aspect, and map the glacial flow velocity. Theprojection mode is UTM zone 42.

3. Method

1) InSAR principleInterferometry processes two radar images to obtain

the phase difference caused by glacier surface motion,and it then calculates glacier displacement within theacquisition time interval between two images accordingto the phase difference (Joughin et al., 1995). The twoSLC radar images are referred to as the master imageand slave image, respectively. The maximum cross-correlation algorithm is used to perform imageregistration on the two images pixel-by-pixel, whichcorrects the antenna gain pattern and nonzero Dopplercentre, thereby correcting Doppler frequency changesalong the orbit caused by side-looking (Joughin et al.,1996). The results of interferometric processing containamplitude and phase information of the same pixels inthe master and slave image after registration. The noisegenerated by the curvature of the earth is removed withprecise orbital information. The SRTM DEM in thestudy area is used to remove terrain phase, and after theremoval of terrain signals, the phase produced by the

surface change is unwrapped from the repeatedlyobserved radar images to obtain the interferogram ofglacier surface deformation (Joughin et al., 1998;Massonnet et al., 1993; Muskett et al., 2008; Scharrooand Visser, 1998).

Because of its high sensitivity to glacier surfacedeformation, InSAR is often used to measure glaciervelocity fields (Joughin et al., 2010). InSAR requiresan effective interference process to obtain accurate flowvelocities (Gourmelen et al., 2011). Feature tracking isalso very widely used to exploit glacier motion andmass balance change, and it is more effective thanInSAR in areas where coherence is not maintained suchas fast-flowing glaciers. The drawback of the featuretracking, compared to InSAR, is that the spatialresolution (and accuracy) for estimating displacementsis lower than InSAR. InSAR techniques have beenwidely used in velocity estimations of glaciers andpolar ice sheets (Joughin et al., 2004) and applied toresearch on surface deformation caused by landsubsidence, earthquakes and volcanic events (Cigna etal., 2012; Massonnet et al., 1993). COSMO-SkyMeddata have very short time baseline (1 day) andPerpendicular baseline (51 m) (Table 1), and theircoherence is effectively maitained, therefore InSARmethod is used in this study to estimate thedisplacement and velocity with higher accuracy.

2) Glacier velocity calculationWe used Gamma software for processing the

interferometric pairs listed in Table 1. Firstly, SLCimages from level-0 raw SAR data sets are generated,then an interferogram from a pair of sub pixelcoregistered SLC images is produced, and the next stepis extraction of deformation pattern from interferogram


– 18 –

Table 1. COSMO-SkyMed data used for InSAR processing

Sensor Acquisition data Orbit type Polarization Time baseline Perpendicular baselineCSKS2 29 January 2010 Descending HH

1 day 51 mCSKS3 30 January 2010 Descending HH


after removing the topographic phase using existingSRTM DEM. During the process of removing thetopographic phase, it is hard to coregister accurately thereference phase to the coordinate system of theinterferogram, and the offsets between the referencephase and the interferogram will lead to phase error.This error and the errors in SRTM DEM will furtherresult in residual errors in extracting glacier flowpatterns. For the purpose of removing most of thetopographic phase with the preliminary DEM, aniterative computation to subtract the topography phasewas implemented. We chose an ice-free and flat areaas the ‘threshold phase region’ and set the value of thethreshold phase to approximately zero, and manuallyset the initial offsets in range and azimuth direction inthe range -10 to 10 pixels, in order to compute andsubtract the reference phase from the interferogramiteratively with different offsets in range and azimuthdirection until the phase of the ice-free area meets thethreshold (Zhou et al., 2011). After this processing,most of the topography phase has been removed from the interferogram. The interferogram, which thetopographic phase is subtracted cleanly, was unwrappedusing algorithms provided by Gamma software andconverted the phase to line-of-sight (LOS) velocity ofthe glacier. Finally, the nunataks near the glaciers wereused as a zero motion reference to calibrate the LOSvelocity.

From the InSAR processing mentioned above, themeasured LOS velocity is merely the surface velocityprojected onto the radar line of sight. To obtain theactual glacier surface flow velocity, four assumptionsare required (Li et al., 2008): 1) glacier moves inparallel to the direction of surface features; 2) the DEM used to remove terrain phase can accuratelyreflect the geomorphological features of groundsurface; 3) elevation change between image pairs usedfor interferometric processing can be neglected; and 4)effect of the atmosphere is negligible. Based on theseassumptions and the methods proposed by Li et al.

(2008) and Kumar et al. (2011), the LOS velocitymeasurements were projected to the glacier surface toget the actual glacier flow velocity (3D).

3) Estimation of equilibrium line altitudeThe glacier mass equilibrium line altitude (ELA)

refers to the altitude of the line connecting points withmass accumulation equal to the melting of glaciers in abalance year on a non-shaded plane (altitude of thelower limit of the additional ice belt) (Cui and Wang,2013). The AAR method is chosen to estimate the massequilibrium line of the five glaciers, and it assumes thatwhen the glacier is in a steady state, the area of theglacier accumulation zone accounts for a fixed ratio ofthe total glacier area at a certain altitude, and the ratiois the AAR value, whereas the altitude and locationindicate the glacier mass equilibrium line (Meier,1962). The method of selecting an appropriate AARvalue is the key step to estimate ELA by the AARmethod. According to the emprical formula (1)proposed by Kern and László (2010), AAR can becalculated.

AAR = 0.0648 × ln S + 0.483 (1)

AAR = (2)

where S is the total area of the glacier, and SELA is thearea of the steady accumulation zone above the glaciermass equilibrium line. At the same time, based on thecalculated emprical AAR value, SELA can be calculatedwith formula (2) according to AAR’s definition byMeier (1962). Thus, the ELA of glacier mass and itslocation can be estimated in support of SRTM DEM.

4) Error estimationIn glacier dynamics research, it is relatively rare to

map the error field of individual glacier, whereas theerrors of each pixel point on the glacier surface are adirect reflection of the accuracy of flow velocitycalculation. The glacier surface velocity error field ΔD is estimated according to the following formulae

SELAS


– 19 –


(Li et al., 2008). The partial derivatives of the actualdisplacement D with respect to ΔL and α are denotedas follows:

ΔD ≈ × ΔL + × Δα (3)

where D is the actual displacement (3 dimensiondisplacement), L the horizontal displacement, α theangle between D and L. The sources of the velocityfield error ΔD include three aspects: SAR imageregistration (ΔD1), geographic position (ΔD2), andsystem errors (ΔD3) (Wuite et al., 2009). Therefore, ΔDis further decomposed into three parts: ΔD1, ΔD2 andΔD3.

ΔD ≈ + D × tan α × Δα = ΔD1 + ΔD2 + ΔD3

(4)

ΔD1 = , ΔD2 = , ΔD3 = D × tan α × Δα (5)

where λ is the wavelength. Because the terrain phasehas been removed effectively through the interferometricprocessing of SAR images using Gamma software, theerror generated by interpreting the image registration(ΔD1) can be regarded as 0.

The error of geographic position (ΔD2) mainlyoriginates from estimations of the vertical baselinelength (B⊥) of the observation system and relativeheight (Δh) between the DEM and the real elevation,i.e., the DEM error (Zebker et al., 1994):

ΔL2 = (6)

where r is the range distance of the COSMO-SkyMedobservation system, and θ is the incidence angle of thesatellite. The DEM error results in noise streaks whensimulating the terrain phase. Therefore, the DEMaccuracy should be considered in the error analysis(Kumar et al., 2011). In the southeast Tibet, which hasvalleys interleaving among the mountains, Δh abovethe altitude of 1750 m is 16 m (Arendt et al., 2002).

ΔD3 is the error generated when calculating the anglebetween the surface slope and aspect between the

displacement along the LOS and actual displacement.Differences in the geometric characteristics of the orbitbetween repeatedly observed image pairs result inchanges in the relative azimuth angle of the entireobservation system (Li et al., 2008). In addition,changes in the satellite orbit cause mismatches of therepeated image pairs in different directions, and thesemismatches also generate registration errors duringInSAR data processing (Strozzi et al., 2001; Kumar etal., 2011). In this study, a single image pair is chosento estimate the glacier flow velocity, thus, the changeof the relative azimuth angle is 0, i.e. ΔD3 = 0.

4. Results and discussion

1) Glacier velocityThe glacier boundary is identified from the

COSMO-SkyMed radar images by visual interpretationand information from the GLIMS (Global Land IceMeasurements from Space) database (http://www.glims.org/). Glacier motion shows a clear relationshipwith terrain undulation and surface morphologicaldifferences, with altitude changes and mountainlocations on both sides constraining the magnitude anddirection of the overall glacier flow velocity (Fig. 2).For Glaciers No.1, No.2, No.4 and No.5, with a narrowoutflow and the larger drop, glacier flow velocitiesbegin to increase from the starting points (firn basins),and continue to increase along the main flow linetowards the glacier mass equilibrium line, then decreasein the flat, open area in the middle of the glaciers. Inthe glacier tongue, lateral and bottom friction becomebigger because of the overall gravity action of muchglacier mass flowed in from the upper valley, and the flow velocities further gradually decrease. Thedistribution of glacier surface flow velocity fields hasa direct relationship with elevation change. Since theelevation decreases from the northwest to the southeast,

∂D∂L

∂D∂α

λ × ΔL2 × cos α



2 × B⊥ × Δhλ × r × sinθ


– 20 –


the glaciers mostly move along this direction. However,because of morphological changes in the surfacegeometry and inverse decreases of elevation, glacierbranches No.2b, No.2c, No.2d, No.4b and No.4c flowin a southwest to northeast direction (Fig. 2).

Clearly, the flow direction of Glacier No.1 issignificantly affected by the glacier shape and the valleyterrain (Fig. 3). The flow velocities are relatively highin steep slope areas near the foot of mountains, withglacial mass accumulated towards the main flow line.Additionally, the flow velocities are generally high in

glacier accumulation zones and gradually decrease withincreased distance from the glacier mass equilibriumline, and these situations are also observed in the flowvelocity fields of glaciers No.2, No.4 and No.5. Thesurface flow velocity of Glacier No.1 reaches amaximum at the glacier mass equilibrium line. Becauseof disordered surface undulations in valleys, the surfaceflow velocity field is discontinuous (Fig. 3).

It is worth noting that the flow velocity of GlacierNo.3 at 4790 m elevation is 33.8 ma-1, and it decreasesto 8.6 ma-1 at 4730 m. However, the flow velocity


– 21 –

Fig. 2. Flow velocity of glaciers in Parlung Zangbo River basin in January 2010.



– 22 –

Fig. 4. Profiles of glacier flow velocity and elevation along the main flow line and mass equilibrium line altitude (ELA).

Fig. 3. Flow velocity and direction of Glacier No. 1.


progressively increases with decreasing elevation andreaches a maximum of 40 ma-1 at 4500 m. Combinedwith optical remote sensing images of Landsat 5, wefind a glacial lake at the terminal of Glacier No.3. Rapidmass exchange between the lake water and glacieraccelerates the flow velocity at the glacier terminal(Björnsson, 1998), and this situation is inconsistent withthe decreasing trend in the surface flow velocities ofthe other four glaciers as elevation decreasing (Fig. 4).

In January, the average flow velocities of the fiveglaciers are 22.9-98.2 ma-1, and the maximum velocityoccurs for Glacier No.1 at 123.9 ma-1 (Figs. 3, 4). Theaverage flow velocities of glaciers in the adjacent basinof the Yigong Zangbo River during summer in 2007were estimated to be 15-206 ma-1 by Advance LandObserving Satellite (ALOS) Phased Array type L-bandSynthetic Aperture Radar (PALSAR) images andfeature tracking method (Ke et al., 2013). Similarly,based on SAR images, the average flow velocities ofYengisogat Glacier in Karakorum Mountains in 2009were estimated at 110-182 ma-1 (Jiang et al., 2011). Theaverage surface flow velocity of glaciers in the WestKunlun Mountains for the period 2007-2010 was 100ma-1 (Yasuda and Furuya, 2013), and those of theInylchek Glacier in Tianshan Mountains for the period2009-2010 were 91.5-109.5 ma-1 (Neelmeijer et al.,2014). Our results are close to the motion patterns ofvalley glaciers mentioned in the above studies andreflect the overall characteristics of glacier motion inwestern China. Additionally, almost all of the glaciersin this basin belong to maritime glaciers, and their flow

velocities are greater than those of continental glaciersinfluenced by the continental climate during the sameseason (Huang and Sun, 1982).

2) Glacier movement and its physical featuresThe flow velocities are higher in the upstream

accumulation zone but lower at the end of the glaciers(Fig. 4). For four glaciers (except for Glacier No.3), theflow velocity presents a fluctuating and decreasingtrend with decreasing elevation. Overall, although theelevation monotonically decreases along the valleys,the flow velocities along the main flow line do notmonotonously decrease with increasing length of theglaciers. Above the glacier mass equilibrium line,because the speed of accumulation is greater than thatof melting, along with elevation decrease, the glaciervelocity increases and reaches a peak value under theinfluence of gravity. After crossing the glacier massequilibrium line, the flow velocities of the four Glaciersgradually decrease.

The five glaciers have different ELAs, but their AARvalues are all higher than 0.6 (Table 2). Meanwhile, theGlacier No.1 has the largest AAR value, area andlength, and also has the largest flow velocity. Althoughsmall area and short length of Glacier No.4, it hasthe highest ELA of 5038 m because of the terrain, andits maximal and mean velocity are 59.2 ma-1 and 29.4ma-1, respectively. According to the study by Markand Post (2002), all five glaciers are in a stable state.In the southeast Tibet, the valley glaciers with lowlatitudes and high altitudes are controlled by the


– 23 –

Table 2. The physical features of 5 glaciers in the Parlung Zangbo River basin

Glacier No.

Maximal Velocity (ma-1)

Distance from S

(m)

Mean Velocity (ma-1)

Elevation range (m)

Area (km2)

Length (km) AAR ELA

(m)

1 123.9 6748 98.2 3400-4701 18.6 18.7 0.6724 43292 76.0 2039 26.0 3622-4291 13.0 9.5 0.6492 42473 39.4 7496 22.9 4368-5099 10.7 9.2 0.6366 48434 59.2 2164 29.4 4334-5138 14.8 9.8 0.6576 50385 55.5 839 23.9 4244-5122 15.6 11.0 0.6610 4705


strong humid climate of the southwest monsoon, andtheir motion state is jointly affected by the evolutionof geomorphological process and climate change(Chen et al., 2012).

3) Error analysisPrevious error analyses are mostly confined to the

cause of the errors, and the specific velocity error fieldis rarely mapped. Assuming that the true values of allglacier surface flow velocities are normally distributed,and based on the accuracy requirements of statistics, allof the values greater than the maximum glacier surfaceflow velocities by one-fold standard deviation areconsidered incorrect. The difference between theminimum error value and maximum flow velocity isthe cut-off value for selection; thus, it is the maximumerror distribution.

As shown in the error fields of Glaciers No.1, No.4and No.5 (Fig. 5), the maximum errors for the threeglaciers are 5.70 ma-1, 3.32 ma-1 and 2.55 ma-1,respectively, and relatively large errors occur in themass accumulation zone at the centre of the glaciers(Because of the complicated terrain, the error fields

of Glaciers No.2 and No.3 cannot be calculated).Combined with the glacier shape and error distribution,we find that the length and width of the glaciers arerelatively uniform, the flow direction is consistent, andthe error distribution is more continuous, which issimilar to branch glacier No.1a and No.4a. However,the error distribution is relatively discrete andfragmented as shown in Glacier No.5. The errors arerelatively large in areas inconsistent with the main flowdirection, such as branch glacier No.4b, No.4c andNo.5b. In January, displacement of discontinuousregions occurs in small areas and on short time scalesin the glacier accumulation zone, whose instantaneouschange will cause local glacier motion in any direction(Bartholomaus et al., 2008). Although such randommotion will have a certain impact on the error analysis,this impact only accounts for a small proportion in theentire glacier and can be considered negligible.

5. Conclusions

Based on COSMO-SkyMed X-band data, relativelyaccurate maps of surface flow velocity fields for fiveglaciers in the Parlung Zangbo River basin are obtainedby InSAR techniques. High-resolution X-band imagepairs can more accurately define the boundary of valleyglaciers and result in high coherence over a short timescale in InSAR processing. The resulting interferencefringes are evident, especially in the cloudy and snowyParlung Zangbo region. In addition, the superiority ofCOSMO-SkyMed data in glacier dynamics researchusing InSAR techniques is illustrated.

The surface morphology and spatial geometriccharacteristics of valleys determine the direction andvelocity of glacier surface flow and further confine theirmotion state. Among the five glaciers in the ParlungZangbo River basin, the maximum flow velocity occursat 123.9 ma-1 in Glacier No.1, whereas the minimumflow velocity occurs at 39.4 ma-1 in Glacier No.3. In


– 24 –

Fig. 5. The error distribution field of glacier No.1, No.4 andNo.5.


comparison, Glacier No.1 has the largest area andlongest length, whereas Glacier No.3 has the smallestarea and shortest length. The average flow velocities ofthe five glaciers range from 22.9 to 98.2 ma-1. Theseglaciers generally move along the direction ofdecreased elevation and present a macroscopicillustration of the motion from the northwest to thesoutheast, with different trends in local areas. The slopeaspect also has important effects on the velocity resultsand accuracy. The error field is relatively fragmentedin areas inconsistent with the main flow line of theglaciers. In greatly undulating areas, such as thesoutheast Tibet, the accuracy of DEM and registrationconditions of DEM-simulated terrain phases has certaineffects on calculations of glacier flow direction andvelocity.

Glacier dynamics research aims to more efficientlyreflect the variation patterns of glacier scale over timeand more accurately extract the flow velocitydistribution at the glacier surface. Mapping glaciervelocity field in local areas is important because it canreflect the sensitivity of specific areas to climate.Current limitations of InSAR measurements overglacier include: (1) a degradation of phase coherencewhen the temporal baseline is larger than a few days,leading to measurement loss and uncertainty in areasof low coherence; and (2) insensitivity to surfacedisplacement in the plane orthogonal to the vector ofthe LOS, resulting in the incapacity to detect motion inthe direction of the satellite ground-track (Gourmelenet al., 2011). However, feature tracking and MultipleAperture InSAR (MAI) can partially overcome theseproblems, therefore, we can take full advantages ofInSAR, feature tracking and MAI, and overcome theirlimitations to get a combining result of actual glaciervelocity with higher accuracy in the future work.

Acknowledgements

This research was supported by the National NatureScience Foundation of China (No. 41830105) and also funded by the International Scholar ExchangeFellowship (ISEF) program at KFAS (KoreanFoundation of Advanced Studies). We stronglyacknowledge European Space Agency (ESA) &National Remote Sensing Center of China (NRSCC,MOST) for providing COSMO-SkyMed SAR dataunder Dragon 3 project (ID: 10612). We would thankthe two anonymous reviewers and editors for valuablecomments and suggestions to greatly improve thepaper.

References

Arendt, A.A., K.A. Echelmeyer, W.D. Harrison, C.S.Lingle, and V.B. Valentine, 2002. Rapid wastageof Alaska glaciers and their contribution torising sea level, Science, 297(5580): 382-386.

Barry, R.G., 2006. The status of research on glaciersand global glacier recession: a review, PhysicalGeography, 30(3): 285-306.

Bartholomaus, T.C., R.S. Anderson, and S.P. Anderson,2008. Response of glacier basal motion totransient water storage, Nature Geoscience, 1:33-37.

Berthier, E., H. Vadon, D. Baratoux, Y. Arnaud, C.Vicent, K.L. Feigl, F. Remy, and B. Legresy,2005. Surface motion of mountain glaciersderived from satellite optical imagery, RemoteSensing of Environment, 95(1): 14-28.

Björnsson, H., 1998. Hydrological characteristics of thedrainage system beneath a surging glacier,Nature, 395: 771-774.

Chen, R.R., S.Z. Zhou, and Y.L. Deng, 2012.Morphological characteristics of glacialdeposits during the last glaciaton-taking the


– 25 –


Parlung Zangbo River Basins as an example,Journal of Glaciology and Geocryology, 34(4):836-847 (in Chinese).

Ciappa, A., L. Pietranera, and F. Battazza, 2010. PeritoMoreno Glacier (Argentina) flow estimation byCOSMO SkyMed sequence of high-resolutionSAR-X imagery, Remote Sensing andEnvironment, 9(15): 2088-2096.

Cigna, F., B. Osmanoglu, E. Cabral-Cano, T.H. Dixon,J.A. Avila-Olivera, V.H. Garduno-Monroy, C.DeMets, and S. Wdowinski, 2012. Monitoringland subsidence and its induced geologicalhazard with Synthetic Aperture RadarInterferometry: A case study in Morelia,Mexico, Remote Sensing of Environment, 117:146-161.

COSMO SkyMed System, 2010. Description and Userguide, http://www.e-geos.it/products/cosmo.html, Accessed on Jun. 7, 2018.

Crosetto, M., O. Monserrat, M. Cuevas-Gonzalez, N.Devanthery, G. Luzi, and B. Crippa, 2015.Measuring thermal expansion using X-bandpersistent scatterer interferometry, ISPRSJournal of Photogrammetry Remote Sensing,100: 84-91.

Cui, H. and J. Wang, 2013. The methods for estimatingthe equilibrium line altitudes of a glacier,Journal of Glaciology and Geocryology, 35(2):345-354 (in Chinese).

Diolaiuti, G., D. Bocchiola, C. D’agata, and C. Smiralia,2012. Evidence of climate change impacuponglaciers’ recession within the Italian, Theoreticaladd Applied Climatology, 109(3-4): 429-445.

Dowdeswell, J.A., B. Unwin, A.M. Nuttall, and D.J.Wingham, 1999. Velocity structure, flowinstability and mass flux on large Arctic ice capfrom satellite radar interferometry, Earth andPlanetary Science Letter, 167(3-4): 131-140.

Goldstein, R.M., H. Engelhardt, B. Kamb, and R.M.Frolich, 1993. Satellite radar interferometry for

monitoring ice sheet motion application to anAntarctic ice stream, Science, 262: 1525-1530.

Gourmelen, N., S.W. Kim, A. Shepherd, J.W. Park,A.V. Sundal, H. Björnsson, and F. Pálsson, 2011.Ice velocity determined using conventionalandmultiple-aperture InSAR, Earth and PlanetaryScience Letter, 307(1-2): 156-160.

Huai, B.J., Z.Q. Li, and S.J. Wang, 2014. Remotesensing analysis of glaciers change in the HeiheRiver Basin, Northwest China, during the recentdecades, Journal of Geographical Sciences,24(6): 993-1008.

Huang, M. and Z. Sun, 1982. Some of flowcharacteristics of continental-type glaciers inChina, Journal of Glaciology and Geocryology,4(2): 35-45 (in Chinese).

Jiang, Z.L., S.Y. Liu, and H.D. Han, 2011. Analyzingmountain glacier surface velocities using SAR Data, Remote Sensing Technology andApplication, 26(5): 640-646 (in Chinese).

Jiang, Z.L., S.Y. Liu, J. Peters, J. Lin, S.C. Long, Y.S.Han, and X. Wang, 2012. Analyzing YengisogatGlacier surface velocities with ALOS PALSARdata feature tracking, Karakoram, China,Environmental Earth Sciences, 67(4): 1033-1043.

Joughin, I.R., 2002. Ice-sheet velocity mapping: acombined interferometric and speckle-trackingapproach, Annals of Glaciology, 34: 195-201.

Joughin, I.R., B.E. Smith, and W. Abdalati, 2010.Glaciological advances made with interferometricsynthetic aperture radar, Journal of Glaciology,56(200): 1026-1042.

Joughin, I.R., D.P. Winebrenner, and M.A. Fahnestock,1995. Observations of ice-sheet motion inGreenland using satellite radar interferometry,Geophysical Research Letters, 22(5): 571-574.

Joughin, I.R., R. Kwok, and M.A. Fahnestock, 1996.Estimation of ice sheet motion using satelliteradar interferometry: method and error analysiswith application to Humboldt Glacier, Greenland,


– 26 –


Journal of Glaciology, 42(142): 564-574.Joughin, I.R., R. Kwok, and M.A. Fahnestock, 1998.

Interferometric estimation of threedimensionalice-flow using ascending and de-scendingpasses, IEEE Transactions on Geoscience andRemote Sensing, 36(1): 25-37.

Joughin, I.R., W. Abdalati, and M. Fahnestock, 2004.Large fluctuations in speed on Greenland’sJakobshavn Isbrae Glacier, Nature, 432: 608-610.

Kang, S., Y. Xu, Q. You, W.A. Flugel, N. Pepin, andT.D. Yao, 2010. Review of climate andcryospheric change in the Tibetan Plateau,Environmental Research Letter, 5(1): 015101.

Ke, C.Q., C. Kou, R. Ludwig, and X. Qin, 2013.Glacier velocity measurements in the easternYigong Zangbo basin, Tibet, China, Journal ofGlaciology, 59(218): 1060-1068.

Kern, Z. and P. Laszlo, 2010. Size specific steady-state accumulation-area ratio: an improvementfor equilibrium-line estimation of smallpalaeoglaciers, Quaternary Science Reviews,29: 2781-2787.

Kumar, V., G. Venkataramana, and K.A. Hogda, 2011.Glacier surface velocity estimation using SAR interferometry technique applyingascending and descending passes in Himalayas,International Journal of Applied EarthObservation and Geoinformation, 13(4): 545-551.

Li, J.J., B.X. Zheng, and X.J. Yang, 1986. Glaciers ofXizang (Tibet), Science Press, Beiging, China.

Li, S., C. Benson, R. Gens, and C. Lingle, 2008. Motionpatterns of Nabesna Glacier (Alaska) revealedby interferometric SAR techniques, RemoteSensing of Environment, 112(9): 3628-3638.

Liu, G.X., H. Jia, R. Zhang, H. Zhang, B. Yu, and M. Sang, 2011. Exploration of subsidenceestimation by persistent scatterer InSAR ontime series of high resolution TerraSAR-X

images, IEEE Journal of Selected Topics inApplied Earth Observations and RemoteSensing, 4(1): 159-170.

Liu, S.Y., X. Qin, and T. Zhang, 2012. Variation of the Ningchan River Glacier No. 3 in theLenglongling Range, East Qilian Mountains,Journal of Glaciology and Geocryology, 34(5):1031-1036 (in Chinese).

Mark, F.M. and A.S. Post, 2002. Recent variations inmass net budgets of glaciers in western northAmerica, International Journal of Climatology,22(1): 107-120.

Massonnet, D., M. Rossi, C. Carmona, F. Adragna, G.Peltzer, K. Feigl, and T. Rabaute, 1993. Thedisplacement field of the Landers earthquakemapped by radar interferometry, Nature, 364(8):138-142.

McCarthy, J.J., O.F. Canziani, and N.A. Leary, 2011.Contribution of Working Group II to the ThirdAssessment Report of the IntergovernmentalPanel on Climate Change (IPCC), Universityof Cambridge Press, Cambridge, UK.

Meier, M.F., 1962. Proposed definitions for glaciermass buget terms, Journal of Glaciology,4(33): 252-263.

Michel, R. and E. Rignot, 1999. Flow of GlaciarMoreno, Argentina, from repeat-pass shuttleimaging radar images: comparison of the phasecorrelation method with radar interferometry,Journal of Glaciology, 45(149): 93-100.

Mohr, J.J., N. Reeh, and S.N. Madsen, 1998. Three-dimensional glacier flow and surface elevationmeasured with radar interferometry, Nature,391: 273-276.

Muskett, R.R., C.S. Lingle, J.M. Sauber, B.T. Rabus,and W. V. Tangborn, 2008. Acceleration ofsurface lowering on the tidewater glaciers ofIcy Bay, Alaska, U.S.A. from InSAR DEMsand ICESat altimetry, Earth Planetary ScienceLetters, 265(3-4): 345-359.


– 27 –


Neelmeijer, J., M. Mahdi, and H. U. Wetzel, 2014.Estimating spatial and temporal variability insurface kinematics of the Inylchek Glacier,Central Asia, using TerraSAR-X data, RemoteSensing, 6(10): 9239-9259.

Scharroo, R. and P. Visser, 1998. Precise orbitdetermination and gravity field improvementfor the ERS satellites, Journal of GeophysicalResearch, 103: 8113-8127.

Schubert, A., A. Faes, A. Kääb, and E. Meier, 2013.Glacier surface velocity estimation using repeatTerraSAR-X images: Wavelet- vs. correlation-based image matching, ISPRS Journal ofPhotogrammetry Remote Sensing, 82: 49-62.

Strozzi, T., U. Wegmueller, L. Tosi, G. Bitelli, and V.Spreckels, 2001. Land subsidence monitoringwith differential SAR interferometry,Photogrammetric Engineering and RemoteSensing, 67(11): 1261-1270.

Wuite, J., K.C. Jezek, X. Wu, K. Farness, and R.Carande, 2009. The velocity field and flowregime of David Glacier and DrygalskiIceTongue, Antarctica, Polar Geography, 32(3-4):111-127.

Yan, S.Y., H.D. Guo, G. Liu, and W. Fu, 2013.Monitoring Muztagh Kuksai Glacier surfacevelocity with L-band SAR data in southwesternXinjiang, China, Environmental Earth Science,70(7): 3175-3184.

Yang, W., T.D. Yao, B.Q. Xu, L.L. Ma, Z.H. Wang, andM. Wan, 2010. Characteristics of recent

temperat glacier fluctuations in the ParlangZangbo River basin, southeast Tibetan Plateau,Chinese Science Bulletin, 55(18): 1775-1780.

Yao, T. D., L. Thompson, W. Yang, W. Yu, Y. Gao, X.Guo, X. Yang, K. Duan, H. Zhao, B. Xu, J. Pu,A. Lu, Y. Xiang, D.B. Kattel, and D. Joswiak,2012. Different glacier status withatmosphericcirculations in Tibetan Plateau and surroundings,Nature Climate Change, 2(9): 663-667.

Yasuda, T. and M. Furuya, 2013. Short-term glaciervelocity changes at West Kunlun Shan,Northwest Tibet, detected by SyntheticAperture Radar data, Remote Sensing ofEnvironment, 128: 87-106.

Zebker, H.A., C.L.Werner, P.A. Rosen, and S. Hensley,1994. Accuracy of topographic maps derivedfrom ERS-1 interferometric radar, IEEETransactions on Geoscience and RemoteSensing, 32(4): 823-836.

Zeng, Q.L., Z.F. Yang, and G.X. Yuan, 2007. SongzongLake: an ice-dammed lake of last glacialmaxmum in Purlung Tasangpo River, SoutheastTibet, Quaternary Sciences, 27(1): 85-92 (inChinese).

Zhou, J.M., Z. Li, X.W. Li, S.Y. Liu, Q. Chen, C. Xie,and B.S. Tian, 2011. Movement estimate of theDongkemadi Glacier on the Qinghai-TibetanPlateau using L-band and C-band spaceborneSAR data, International Journal of RemoteSensing, 32(22): 6911-6928.


– 28 –


Documents

InSAR-based Glacier Velocity Mapping in the Parlung Zangbo …sar.kangwon.ac.kr/paper/kjrs_kecq.pdf · 2019. 2. 28. · abundant rainfall at ~1000-3000 mm above the glacier mass equilibrium