Inter-annual surface evolution of an Antarctic blue-ice moraine using

29 Pages·2017·2.14 MB·English

Edinburgh Research Explorer

Discussion Paper: Inter-annual surface evolution of an Antarctic

blue-ice moraine using multi-temporal DEMs

Citation for published version:
Westoby, MJ, Dunning, SA, Woodward, J, Hein, A, Marrero, S, Winter, K & Sugden, D 2015, 'Discussion

Paper: Inter-annual surface evolution of an Antarctic blue-ice moraine using multi-temporal DEMs', Earth
Surface Dynamics Discussions. https://doi.org/10.5194/esurfd-3-1317-2015

Digital Object Identifier (DOI):

10.5194/esurfd-3-1317-2015

Link:
Link to publication record in Edinburgh Research Explorer

Document Version:
Publisher's PDF, also known as Version of record

Published In:

Earth Surface Dynamics Discussions

Publisher Rights Statement:

© Author(s) 2015. This work is distributed
under the Creative Commons Attribution 3.0 License.

General rights

Copyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s)
and / or other copyright owners and it is a condition of accessing these publications that users recognise and
abide by the legal requirements associated with these rights.

Take down policy
The University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorer
content complies with UK legislation. If you believe that the public display of this file breaches copyright please
contact [email protected] providing details, and we will remove access to the work immediately and
investigate your claim.

Download date: 30. Jan. 2023
EarthSurf.Dynam.Discuss.,3,1317–1344,2015 D
www.earth-surf-dynam-discuss.net/3/1317/2015/ isc ESURFD
u
doi:10.5194/esurfd-3-1317-2015 s
s
©Author(s)2015.CCAttribution3.0License. io 3,1317–1344,2015
n
P
a
Thisdiscussionpaperis/hasbeenunderreviewforthejournalEarthSurfaceDynamics(ESurfD). p
e Inter-annual surface
PleaserefertothecorrespondingfinalpaperinESurfifavailable. r
evolution of an
|
Antarctic blue-ice
Inter-annual surface evolution of an D
moraine
is
c
u
Antarctic blue-ice moraine using s M.J.Westobyetal.
s
io
n
multi-temporal DEMs P
a
p
e TitlePage
r
1 2 1 3 3
M. J. Westoby , S. A. Dunning , J. Woodward , A. S. Hein , S. M. Marrero , | Abstract Introduction
1 3
K. Winter , and D. E. Sugden
D Conclusions References
is
1DepartmentofGeography,EngineeringandEnvironment,NorthumbriaUniversity, cus Tables Figures
NewcastleuponTyne,UK s
2SchoolofGeography,PoliticsandSociology,NewcastleUniversity,NewcastleuponTyne,UK ion
3SchoolofGeoSciences,UniversityofEdinburgh,Edinburgh,UK Pa J I
p
e
Received:29October2015–Accepted:3November2015–Published:18November2015 r J I
| Back Close
Correspondenceto:M.J.Westoby([email protected])
D
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion. is FullScreen/Esc
c
u
s
sio Printer-friendlyVersion
n
P InteractiveDiscussion
a
p
e
r
1317 |
Abstract D
is
c ESURFD
u
Multi-temporal and fine resolution topographic data products are being increasingly s
s
usedtoquantifysurfaceelevationchangeinglacialenvironments.Inthisstudy,weem- io 3,1317–1344,2015
n
ploy3-Ddigitalelevationmodel(DEM)di◆erencingtoquantifythetopographicevolution P
a
p
5 of a blue-ice moraine complex in front of Patriot Hills, Heritage Range, Antarctica. Ter- e Inter-annual surface
r
restrial laser scanning (TLS) was used to acquire multiple topographic datasets of the evolution of an
|
moraine surface at the beginning and end of the austral summer season in 2012/2013 Antarctic blue-ice
and during a resurvey field campaign in 2014. A complementary topographic dataset D moraine
is
was acquired at the end of season 1 through the application of Structure-from-Motion c
u
10 (SfM) photogrammetry to a set of aerial photographs taken from an unmanned aerial ssio M.J.Westobyetal.
vehicle (UAV). Three-dimensional cloud-to-cloud di◆erencing was undertaken using n
P
theMultiscaleModeltoModelCloudComparison(M3C2)algorithm.DEMdi◆erencing a
p
revealed net uplift and lateral movement of the moraine crests within season 1 (mean e TitlePage
r
uplift 0.10m), with lowering of a similar magnitude in some inter-moraine depres-
⇠ | Abstract Introduction
sions and close to the current ice margin. Our results indicate net uplift across the site
15
between seasons 1 and 2 (mean 0.07m). This research demonstrates that it is possi- Dis Conclusions References
c
bletodetectdynamicsurfacetopographicalchangeacrossglacialmorainesovershort us Tables Figures
(annual to intra-annual) timescales through the acquisition and di◆erencing of fine- s
io
resolution topographic datasets. Such data o◆er new opportunities to understand the n
P J I
process linkages between surface ablation, ice flow, and debris supply within moraine a
20 p
e
ice. r J I
| Back Close
1 Introduction Dis FullScreen/Esc
c
u
Fine-resolution topographic data products are now routinely used for the geomorpho- s
sio Printer-friendlyVersion
metriccharacterisationofEarthsurfacelandforms(e.g.Passalacquaetal.,2014,2015; n
Tarolli, 2014). Recent decades have seen the advent and uptake of a range of survey- P InteractiveDiscussion
25 a
p
ing technologies for characterising the form and evolution of Earth surface topography e
r
1318 |
at the macro- (landscape; km), meso- (landform; m) and micro-scales (patch-scale; D
is
cm–mm). These technologies have included, amongst others, the use of satellite re- c ESURFD
u
s
motesensingtechniques(e.g.Kääb,2002;Smithetal.,2006;Farretal.,2007;Stumpf, s
io 3,1317–1344,2015
2014; Noh and Howat, 2015), as well as field-based surveying platforms such as elec- n
P
5 tronic distance meters (total station; e.g. Keim et al., 1999; Fuller et al., 2003), di◆er- a
p
ential global positioning systems (dGPS; e.g. Brasington et al., 2000; Wheaton et al., e Inter-annual surface
r
2010), terrestrial laser scanning (TLS; e.g. Rosser et al., 2005; Hodge et al., 2009), evolution of an
|
airborne light detection and ranging (LiDAR; e.g. Bollmann et al., 2011) and softcopy Antarctic blue-ice
D
or digital photogrammetry (e.g. Micheletti et al., 2015). moraine
is
c
10 More recently, geoscientists are increasingly adopting low-cost Structure-from- u
s M.J.Westobyetal.
Motion with multi-view stereo (SfM-MVS) methods, which employ computer vision and s
io
multi-view photogrammetry techniques to recover surface topography using optical n
P
(e.g.James andRobson,2012; Westobyetal., 2012;Javernicket al.,2014;Micheletti a
p
e TitlePage
et al., 2014; Woodget et al., 2015; Smith and Vericat, 2015) or thermal imagery (e.g. r
15 Lewis et al., 2015). Concomitant developments in lightweight unmanned aerial vehicle | Abstract Introduction
(UAV) technology, specifically decreasing system costs, increased portability, and im-
D Conclusions References
provements in the accessibility of flight planning software have encouraged the acqui- is
c
sition of repeat, fine-resolution (m to cm) topographic data products from low-altitude us Tables Figures
s
aerial photography platforms (e.g. Niethammer et al., 2010; Ouédraogo et al., 2014). io
n
Furthermore,thedi◆erencingoftopographicdatasetsacquiredatdi◆erenttimesisnow
20 P J I
a
an established method for quantifying the transfer of mass and energy through land- p
e
scapes at the spatial scales of observation at which many processes operate (Pas- r J I
salacqua et al., 2015). | Back Close
To date, fine-resolution topographic datasets produced using airborne or ground-
D
based light detection and ranging (LiDAR), or terrestrial or low-altitude aerial digital is FullScreen/Esc
25 c
u
photogrammetry have been used for a diverse range of applications in various glacial, s
proglacial, and periglacial environments at a range of scales, including: the quantifica- sio Printer-friendlyVersion
n
tion of ice surface evolution (e.g. Baltsavias et al., 2001; Pitkänen and Kajuutti, 2004; P InteractiveDiscussion
a
Keutterling and Thomas, 2006; Schwalbe and Maas, 2009; Immerzeel et al., 2014; p
e
r
1319 |
Pepin et al., 2014; Whitehead et al., 2014; Gabbud et al., 2015; Kraaijenbrink et al., D
is
2015;Piermatteietal.,2015;Ryanetal.,2015);mappingtheredistributionofproglacial c ESURFD
u
s
sediment(e.g.Milanetal.,2007;Irvine-Fynnetal.,2011;Dunningetal.,2013;Staines s
io 3,1317–1344,2015
et al., 2015) and moraine development (Chandler et al., 2015); the characterisation of n
P
5 glacier surface roughness (e.g. Sanz-Ablanedo et al., 2012; Irvine-Fynn et al., 2014), a
p
sedimentology (Westoby et al., 2015), and hydrology (Rippin et al., 2015); as well as e Inter-annual surface
r
input data for surface energy balance modelling (e.g. Arnold et al., 2006; Reid et al., evolution of an
|
2012); and for characterising glacial landforms in formerly glaciated landscapes (e.g. Antarctic blue-ice
D
Smith et al., 2009; Tonkin et al., 2014; Hardt et al., 2015). moraine
is
c
10 In this study, we utilise fine-resolution topographic datasets to quantify the surface u
s M.J.Westobyetal.
evolution of a blue-ice moraine complex in a remote part of Antarctica. Blue-ice areas s
io
cover approximately 1% of Antarctica’s surface area (Bintanja, 1999), yet they remain n
P
relativelyunderstudied.Relictblue-icemorainespreservedonnunataksarekeyindica- a
p
e TitlePage
torsoficesheetelevationchanges;however,limiteddataexistonratesandpatternsof r
15 surfacereorganisation,whichmaybeofuseforcontextualisingtheresultsof,forexam- | Abstract Introduction
ple, cosmogenic nuclide dating and geomorphological mapping. This research seeks
D Conclusions References
to quantify the short-term surface evolution of a moraine complex in Patriot Hills, Her- is
c
itageRange,Antarctica(Fig.1),throughthedi◆erencingandanalysisofmulti-temporal us Tables Figures
s
topographic datasets acquired using TLS and the application of SfM-MVS photogram- io
n
metry to optical imagery acquired from a low-altitude UAV sortie.
20 P J I
a
p
e
r J I
2 Study site
| Back Close
The study site is a blue-ice moraine complex, located on the northern flank of the Dis FullScreen/Esc
Patriot Hills massif at the southern-most extent of Heritage Range, West Antarctica c
u
s
(Fig. 1). Blue-ice moraine formation is hypothesised to be the result of preferential sio Printer-friendlyVersion
ablation of marginal ice by katabatic winds, which in turns prompts the modification n
25
P InteractiveDiscussion
of ice flow and englacial sediment transport pathways such that basal sediment is a
p
brought to the ice surface, where it is deposited (e.g. Bintanja, 1999; Sinisalo and e
r
1320 |
Moore, 2010; Fogwill et al., 2012; Spaulding et al., 2012). The site comprises a series D
is
ofbroadlyeast–westorientedmoraineridgesandinter-morainetroughs,aswellasan c ESURFD
u
s
area of subdued moraine topography immediately adjacent to the ice margin. At this s
io 3,1317–1344,2015
location, the active blue-ice moraines occupy an altitudinal range of 60–70m above n
P
5 the ice margin ( 730ma.s.l.), and extend for a distance of up to 350m into a bedrock a
⇠ p
embayment. The blue-ice moraines can be traced for a distance of >4km to the east e Inter-annual surface
r
andnorth-east,paralleltotherangefront,andfillice-marginalembayments.Thesiteis evolution of an
|
geomorphologically and sedimentologically complex (e.g. Vieira et al., 2012; Westoby Antarctic blue-ice
D
et al., 2015), and, along with moraine ridges and troughs, includes areas of subdued moraine
is
c
10 ice-marginal topography with thermokarst melt ponds, local gullying and crevassing u
s M.J.Westobyetal.
on ice-proximal and distal moraine flanks, as well as solifluction deposits at the base s
io
of the surrounding hillslopes. The bedrock hillslopes are overlain by a till drape with n
P
rare, large exotic sandstone boulder erratics which have some evidence of periglacial a
p
e TitlePage
reworking.Fieldobservationssuggestthattheblue-icemorainesaredynamicfeatures r
15 which are undergoing localised surface changes. It is these short-term changes which | Abstract Introduction
are the subject of investigation in this paper.
D Conclusions References
is
c
us Tables Figures
3 Methods and data products s
io
n
This research employs two methods for reconstructing moraine surface topography, P J I
a
p
specifically TLS and SfM-MVS photogrammetry. Two field campaigns at Patriot Hills e
r J I
wereundertakenwitha12monthsurveyinterval.Briefly,TLSdatawereacquiredatthe
20
| Back Close
beginning and end of austral summer season 1 (December 2012 and January 2013,
respectively), and in a short resurvey visit in season 2 (January 2014). Low-altitude Dis FullScreen/Esc
aerial optical photography was acquired from a UAV at the end of season 1 and was c
u
s
usedastheprimaryinputtoSfM-MVSprocessing.Thefollowingsectionsdetailthetwo sio Printer-friendlyVersion
methods of topographic data acquisition, data processing, and subsequent analysis n
25
P InteractiveDiscussion
using “cloud-to-cloud” di◆erencing. a
p
e
r
1321 |
3.1 Topographic data acquisition D
is
c ESURFD
u
3.1.1 Terrestrial laser scanning s
s
io 3,1317–1344,2015
n
TLSdatawereacquiredusingaRieglLMS-Z620time-of-flightlaserscanner,settoac- P
quire ⇠11000pointss�1 in the near-infrared band at horizontal and vertical scanning ape Inter-annual surface
5 incrementsof0.031�,equivalenttoapointspacingof0.05matadistanceof100mand r evolution of an
with a beam divergence of 15mm per 100m. Data were acquired from six locations |
Antarctic blue-ice
across the site at the beginning of season 1 (7–11 December 2012; Fig. 1; Table 1). D
moraine
Two of these positions were re-occupied at the end of season 1 (9 January 2013) and isc
u
threepositionswerereoccupiedinseason2(Fig.1;14January2014).Followingman- s M.J.Westobyetal.
s
ual editing and the automated removal of isolated points to improve data quality, each io
10 n
set of scans were co-registered in Riegl RiSCAN PRO software (v. 1.5.9) using a two- P
a
p
step procedure employing coarse manual point-matching followed by the application e TitlePage
r
of a linear, iterative, least-squares minimisation solution to reduce residual alignment
| Abstract Introduction
error. Individual scans were then merged to produce a single 3-D point cloud for each
scandate.Mergedscandatafromtheendofseasons1and2weresubsequentlyreg- D Conclusions References
15 is
istered to the scan data from the beginning of season 1 using the methods described c
us Tables Figures
above (Table 1). s
io
n
3.1.2 Structure-from-Motion with Multi-View Stereo photogrammetry P J I
a
p
e
r J I
Low-altitude aerial photographs of the study site were acquired using a 10-Megapixel
PanasonicLumixDMC-LX5compactdigitalcamerawithafixedfocallength(8mm)and | Back Close
20
automatic exposure settings, mounted in a fixed, downward-facing (nadir) perspective D
is FullScreen/Esc
on a sub-5kg fixed-wing UAV. Photographs were acquired in a single sortie lasting c
u
5min. A total of 155 photographs were acquired at a 2s interval at an approximate s
⇠ sio Printer-friendlyVersion
ground height of 120m, producing an average image overlap of 80%, and an approxi- n
25 mate ground resolution of 0.07m2 per pixel. Mean point density was 300pointsm�2, Pa InteractiveDiscussion
2 ⇠ p
compared to a mean of 278pointsm� for the TLS datasets. Motion blur of the input e
r
1322 |
imageswasnegligibleduetofavourableimageexposureconditionsandanappropriate D
is
UAV flying height and speed. c ESURFD
u
s
UAV photographs were used as input to SfM reconstruction using the proprietary s
io 3,1317–1344,2015
Agisoft PhotoScan Professional Edition (v. 1.1.6) software. Unique image tie-points n
P
5 which are stable under variations in view perspective and lighting are identified and a
p
matched across input photographs, similar to Lowe’s (2004) Scale Invariant Feature e Inter-annual surface
r
Transform(SIFT)method.Aniterativebundleadjustmentalgorithmisusedtosolvefor evolution of an
|
internal and external camera orientation parameters and produce a sparse 3-D point Antarctic blue-ice
D
cloud. The results of the first-pass camera pose estimation were scrutinised and only moraine
is
c
10 3-Dpointswhichappearinaminimumof3photographsandpossessedareprojection u
s M.J.Westobyetal.
error of <1.0 were retained. A two-phase method of UAV-SfM data registration was s
io
employed:(1)groundcontrolwasobtainedbyidentifyingcommonfeaturesintheUAV- n
P
SfM photographs and TLS data from the end of season 1 (acquired 4 days after the a
p
e TitlePage
SfMdata;Table1),suchasthecornersoflarge,well-resolvedboulders.GCPdatawere r
15 used to optimise the initial camera alignment and transform the regenerated UAV-SfM | Abstract Introduction
datatothesameobjectspaceastheTLSdata,producinganxyzRMSerrorof0.23m.
D Conclusions References
(2) following dense reconstruction, 3-D point data were exported to RiSCAN PRO (v. is
c
1.5.9) software, and a linear, iterative, least-squares minimisation employing surface us Tables Figures
s
plane matching was used to improve the alignment and reduce the xyz RMS error to io
n
0.03m.
20 P J I
a
p
e
3.2 Cloud-to-cloud di◆erencing r J I
| Back Close
Three-dimensional “cloud-to-cloud” distance calculations were used to quantify
D
moraine surface evolution (e.g. Lague et al., 2013). Since the dominant direction of is FullScreen/Esc
c
surface evolution across the study site was unknown a priori, the application of an al- u
s
25 gorithm that is capable of detecting fully three-dimensional topographic change was sio Printer-friendlyVersion
deemed to be the most appropriate method in this context. To this end, we employ the n
P InteractiveDiscussion
Multiscale Model to Model Cloud Comparison (M3C2) algorithm (Lague et al., 2013; a
p
e
r
1323 |
BarnhartandCrosby,2013),implementedintheopen-sourceCloudComparesoftware D
is
(v. 2.6.1) for change detection. c ESURFD
u
s
TheM3C2algorithmimplementstwomainprocessingstepstocalculate3-Dchange s
io 3,1317–1344,2015
between two point clouds: (1) estimation of surface normal orientation at a scale con- n
P
5 sistent with local surface roughness, and (2) quantification of the mean cloud-to-cloud a
p
distance (i.e. surface change) along the normal direction (or orthogonal vector), which e Inter-annual surface
r
includes an explicit calculation of the local confidence interval. A point-specific normal evolution of an
|
vector is calculated by fitting a plane to neighbouring 3-D points that are contained Antarctic blue-ice
D
within a user-specified search radius. To avoid the fluctuation of normal vector orien- moraine
is
c
10 tations and a potential overestimation of the distance between two point clouds, the u
s M.J.Westobyetal.
radius, or scale, used for normal calculation needs to be larger than the topographic s
io
roughness, which is calculated as the standard deviation of local surface elevations n
P
(�). The orientation of the surface normal around a point, i, is therefore dependent a
p
e TitlePage
on the scale at which it is computed (Lague et al., 2013). A trial-and-error approach r
15 was employed to reduce the estimated normal error, Enorm (%), through refinement of | Abstract Introduction
a re-scaled measure of D, ⇠, where:
D Conclusions References
is
D c
⇠(i)= . (1) us Tables Figures
� (D) s
i io
n
Using this re-scaled measure of D, ⇠ can be used as an indicator of estimated normal Pa J I
p
orientationaccuracy,suchthatwhere⇠fallsintherange 20–25,theestimatednormal e
⇠ r J I
errorisE <2%(Lagueetal.,2013).Afixednormalscalingof2mwasfoundtobe
20 norm
| Back Close
sucient to ensure that ⇠> 20 for >98% of points in each topographic dataset.
The radius of the projection cylinder, d, within which the average surface elevation Dis FullScreen/Esc
of each cloud is calculated, was specified as 2m. This scaling ensured that the num- cu
s
ber of points sampled in each cloud was �30, following guidance provided by Lague sio Printer-friendlyVersion
etal.(2013).M3C2executiontook 0.3hforeachdi◆erencingtaskonadesktopcom- n
25
⇠ P InteractiveDiscussion
puteroperatingwith32GBofRAM,anda3.4GHzCPU.Cloud-to-clouddistancesand a
p
statistics were projected onto the original point cloud. M3C2 output was subsequently e
r
1324 |
masked to exclude points where change is lower than level of detection threshold for D
is
a 95% confidence level, LoD95%(d), which is defined as: c ESURFD
u
s
s
�1(d)2 �2(d)2 ion 3,1317–1344,2015
LoD95%(d)=±1.96 n1 + n2 +reg!, (2) Pap
e Inter-annual surface
r
where d is the radius of the projection cylinder, reg is the user-specified registration evolution of an
|
error, for which we substitute the propagated root mean square alignment error for Antarctic blue-ice
5
D
point clouds n1 and n2 (Table 2; Eq. 1) and assume that this error is isotropic and is moraine
c
spatially uniform across the dataset. u
s M.J.Westobyetal.
To calculate the total propagated error for each di◆erencing epoch, � , the esti- s
DoD io
mates of errors in each point cloud (i.e. the sum of the average scan-scan RMS error n
P
and a project-project RMS error, where applicable) were combined using: a
10 p
e TitlePage
r
�DoD= �C2 +�C2 , (3) | Abstract Introduction
1 2
q D Conclusions References
where �2 and �2 are the RMS errors associated with point clouds C and C . is
C1 C2 1 2 cus Tables Figures
s
io
n
4 Short-term topographic evolution of blue-ice moraines
P J I
a
p
Theresultsof3-Dcloud-to-clouddi◆erencingaresummarisedinFigs.3to5.Threshold er J I
levels of change detection ranged from 0.094–0.103m. The upper (i.e. most conser-
15 | Back Close
vative) bound of this range was applied to the results from all di◆erencing epochs, so
D
thatonly3-Dsurfacechangesgreaterthan0.103mwereconsideredinthesubsequent is FullScreen/Esc
c
analysis. The horizontal (xy) and vertical (z) components of 3-D surface change were u
s
separated to aid the analysis and interpretation of moraine surface evolution. Vertical sio Printer-friendlyVersion
n
surfacechangesforarangeofepochs,encompassingintra-annualandannualchange,
20 P InteractiveDiscussion
a
are displayed in Fig. 3, whilst the horizontal component of 3-D change are shown in p
e
Fig. 4. The longest di◆erencing epoch, representing a period of 400 days (Fig. 3b) r
⇠
1325 |