Remote sensing-based quantification and analysis of earth
surface processes in periglacial and glacial environments
– case studies from Iceland, Svalbard and southern Norway
Bjørn Wangensteen
Dr. Scient. Thesis
Oslo 2006
Department of Geosciences
Faculty of Mathematics and Natural Sciences
University of Oslo
Contents
Part I
Overview
ABSTRACT....................................................................................................................................................... VII
ACKNOWLEDGEMENTS................................................................................................................................ IX
1
INTRODUCTION ....................................................................................................................................... 3
1.1
1.2
1.3
2
MOTIVATION ........................................................................................................................................ 3
OBJECTIVES .......................................................................................................................................... 4
OUTLINE ............................................................................................................................................... 5
GEOMORPHIC PROCESSES IN PERIGLACIAL AND GLACIAL ENVIRONMENTS................. 7
2.1
GEOMORPHIC PROCESSES...................................................................................................................... 7
2.2
PERIGLACIAL AND GLACIAL ENVIRONMENTS ........................................................................................ 9
2.2.1
Slope movements and permafrost creep .......................................................................................... 9
2.2.2
Glacier dynamics .......................................................................................................................... 13
2.2.3
Glacier surface change ................................................................................................................. 15
2.2.4
Cryogenic weathering and permafrost coastal cliff retreat .......................................................... 16
3
REMOTE SENSING OF GEOMORPHIC PROCESSES IN PERIGLACIAL AND GLACIAL
ENVIRONMENTS.............................................................................................................................................. 19
3.1
3.1.1
3.1.2
3.1.3
3.1.4
3.1.5
3.2
3.2.1
3.2.2
3.2.3
3.2.4
3.2.5
3.2.6
3.2.7
3.2.8
3.3
3.4
3.4.1
3.4.2
3.4.3
4
DIGITAL PHOTOGRAMMETRY AND OPTICAL REMOTE SENSING ............................................................ 19
Geometry ....................................................................................................................................... 20
Automatic generation of DTMs ..................................................................................................... 21
Orthophotos................................................................................................................................... 24
Mapping earth surface changes by DTM differencing .................................................................. 25
Mapping earth surface displacements from multitemporal optical imagery................................. 26
INTERFEROMETRIC SAR (INSAR) ...................................................................................................... 28
Geometry ....................................................................................................................................... 29
Unwrapping................................................................................................................................... 29
DTMs from InSAR ......................................................................................................................... 31
Differential interferometry (DInSAR)............................................................................................ 31
Limitations..................................................................................................................................... 32
Mapping earth surface changes and displacements by satellite InSAR ........................................ 33
Coherence ..................................................................................................................................... 35
Permanent scatterers technique .................................................................................................... 35
LASER SCANNING ................................................................................................................................ 36
COMPARISON OF THE DIFFERENT TECHNIQUES.................................................................................... 36
DTM differencing and surface elevation changes......................................................................... 36
Measuring earth surface displacements from multi-temporal imagery......................................... 38
Remote sensing techniques in relation to geomorphic processes.................................................. 40
SUMMARY OF RESEARCH .................................................................................................................. 41
4.1
STUDY AREAS ..................................................................................................................................... 41
4.2
METHODS ........................................................................................................................................... 46
4.2.1
Coastal cliff retreat rates and glacier surface elevation change from differencing DTMs ........... 46
4.2.2
Glacier, rock glacier and slope displacements from cross-correlation matching of orthophotos 47
4.2.3
Measuring glacier displacements using height differentiated InSAR scenes ................................ 49
4.3
ANALYSIS AND RESULTS ..................................................................................................................... 51
4.3.1
Coastal cliff retreat rates and glacier surface elevation change from differencing DTMs ........... 51
4.3.2
Glacier, rock glacier and slope displacements from cross-correlation matching of orthophotos 53
4.3.3
Glacier displacements from height differentiated InSAR scenes................................................... 59
5
GENERAL DISCUSSION ........................................................................................................................ 63
5.1
5.2
TECHNIQUE VALIDATION AND APPLICABILITY .................................................................................... 63
GEOMORPHIC PROCESSES AND ANALYSIS ........................................................................................... 66
iii
5.2.1
5.2.2
Flow lines for travel time and age estimates................................................................................. 66
Geomorphic work and retreat rates .............................................................................................. 67
6
CONCLUSIONS........................................................................................................................................ 71
7
OUTLOOK AND RECOMMENDATIONS ........................................................................................... 73
REFERENCES.................................................................................................................................................... 75
iv
Part II
Papers
The thesis is based on four papers. These papers are referred to by their Roman numbers in
the text.
I) Wangensteen, B., Guðmundsson , Á., Eiken, T., Kääb, A., Farbrot, H. and
Etzelmüller, B. Surface displacements and surface age estimates for creeping slope
landforms in northern and eastern Iceland using digital photogrammetry. Accepted for
publication in Geomorphology.
II) Wangensteen B., Eiken, T., Ødegård, R.S. and Sollid, J.L. Measuring coastal cliff
retreat in the Kongsfjorden area, Svalbard using terrestrial photogrammetry. Submitted for
publication in Earth Surface Processes and Landforms.
III) Wangensteen, B., Tønsberg O.M., Kääb, A., Eiken T and Hagen, J.O. 2006. Surface
elevation change and high-resolution surface velocities for advancing outlets of
Jostedalsbreen. Geografiska Annaler. Vol. 88A, 55-74.
IV) Wangensteen, B., Weydahl, D. J., & Hagen, J. O. 2005. Mapping glacier velocities on
Svalbard using ERS tandem DInSAR data. Norsk Geografisk Tidsskrift-Norwegian Journal of
Geography. Vol. 59, 276-285.
v
Abstract
Periglacial environments cover 35 % of the earth’s land area (Williams & Smith 1989)
and glacial environments cover 10% (Paterson 1994). A thorough understanding of the
geomorphic processes that form and change the earth’s surface in these areas is therefore
important, especially in the advent of climate change and expanding human use. Within this
thesis remote-sensing-based techniques are applied to measure and quantify processes that
govern geomorphic development in selected periglacial and glacial environments. The main
objective of the thesis is to address and validate different remote sensing techniques for
geomorphic process quantification in terms of applicability and accuracy. Case studies from
Iceland, Svalbard and southern Norway are presented.
Surface displacements of one debris covered glacier, several rock glaciers and slope
accumulations in northern and eastern Iceland, in addition to three glacier outlets of
Jostedalsbreen in southern Norway, are mapped by cross-correlation matching of multitemporal orthophotos. Flow lines used for age estimates for some of the Icelandic landforms
are made from the surface displacement data.
A method for calculation glacier velocities for three Svalbard glaciers from ERS
tandem height differentiated InSAR scenes is developed. The line of sight glacier velocities
are decomposed into the flow direction of the glacier using the same DTM used for the height
differencing.
Differencing of multitemporal DTMs is done for the three outlets of Jostedalsbreen in
order to investigate surface elevation change between the years 1984, 1993, 1997 and 2001 by
digital photogrammetry.
DTM differencing is also used for measuring coastal cliff retreat rates in the
Kongsfjorden area at Svalbard. Retreat rates of 3.1 and 2.7 mm a-1 are revealed for the two
sites investigated by using terrestrial digital photogrammetry.
vii
Acknowledgements
First of all I want to thank my supervisors Bernd Etzelmüller, Trond Eiken and Rune S.
Ødegård. Bernd Etzelmüller’s endless enthusiasm, encouragement and will to help have been
decisive for me being able to carry out this thesis. Trond Eiken’s knowledge and expertise in
the use of surveying and photogrammetric equipment has, in addition to always being
available for advice and help, been invaluable. I also want to thank Rune S. Ødegård for
helping me with planning and fieldwork for the coastal retreat measurements. Both Trond and
Runes are in large parts responsible for the fieldwork on Svalbard in the summers of 2002 and
2004 being a success.
I want to express my sincere gratitude to Andi Kääb for inviting me to stay at the
Department of Geography at the University of Zürich-Irchel the winter of 2003-2004. Your
help has been very important for the production of half of the papers in this thesis. I also want
to thank all other employees and students in Zürich for making our stay such a pleasant and
successful one.
I want to thank Johan Ludvig Sollid for involving me in the Artic Coastal Dynamics
project, which led to the paper on coastal cliff retreat on Svalbard. Jon Ove Hagen is thanked
for involving me in the Glaciorisk project which financed the air photo campaign utilized for
the results presented in the paper III. I am also grateful to Águst Guðmundsson for initiating
the permafrost project in Iceland together with Bernd Etzelmüller, and for involving me. This
cooperation led to some unforgettable weeks of fieldwork in Iceland in the summer of 2003
and the scientific results presented in Paper I. The help from Dan-Johan Weydahl and Jon
Ove Hagen, who supervised me during the work with my M.Sc-thesis, is also greatly
acknowledged together with the work leading to paper IV within this thesis.
I want to express my gratitude to all the co-authors of the papers presented in this thesis
and to all fellow Ph.D-students, employees and students at the Department of Geosciences
Section of Physical Geography for making my years as a PhD-student such pleasant ones and
for many fun excursions and conference trips. Special thanks are extended to Gaute
Lappegard for acting as a safety valve for many of the frustrations during the work with this
thesis. Sorry for all the interruptions…
This study has been financially supported by the Faculty of Mathematics and Natural
Sciences, University of Oslo, by scholarships from the Norwegian Research Council on
behalf of the Norwegian Polar Committee, by the EU project Glaciorisk (EVG1-2000-00512),
ix
and by INTAS through the project Arctic coasts of Eurasia: dynamics, sediment budget and
carbon flux in connection with permafrost degradation (INTAS-2001-2329). The Norwegian
Research Council is also thanked for financing my stay in Zürich.
I also want to thank my parents for providing a lot of fresh air while growing up and for
encouraging me to pursue my interests; both academic and non-academic.
Above all I want to express my sincere gratitude to my wife Teresia for her support and
help, and especially for your and Urda’s cooperation and understanding during the last
months work with this thesis.
Oslo,
19 March 2006
Bjørn Wangensteen
x
Part I
Overview
1
1 Introduction
1.1
Motivation
Periglacial environments cover 35 % of the earth’s land area (Williams & Smith 1989)
and glacial environments cover 10% (Paterson 1994). A thorough understanding of the
geomorphic processes that form and change the earth’s surface in these areas is therefore
important, especially in the advent of climate change and expanding human use. During the
last two centuries geomorphology has developed from a more descriptive science to a science
focusing on the geomorphic processes and measuring characteristics of those (Bloom 1998).
A process in geomorphology can be described as (a) the chemical alteration of minerals and
the fracturing of solids in to smaller pieces (weathering) and (b) the translocation of solutes
and solids by various agents (erosion, transport and sedimentation). These processes alter the
topography during time. The quantification of these processes is normally only possible in
defined areas of limited size, involving heavy field instrumentation. The results are often
point observations that have to be extrapolated in space and time, with a high degree of
uncertainty. With the emergence of satellite and aerial remote sensing techniques during the
last decades, tools have been developed to quantify rates of geomorphic processes spatially in
terms of translocation of solids and masses (e.g. ice). The advantage of these tools is that they
allow an increasing number of points to be measured within an area, opening for statistically
sound interpolation of velocity and flux fields. These tools are therefore of great importance
for validation of field measurements and derived geomorphic models. With the availability of
images at least back to before World War II in many areas, the time-dependant variation of
surface movements could also be addressed with a higher degree of accuracy. Both these facts
allow for a better and more certain estimation of erosion rates and impact of climate variations
on surface processes.
This study is based on participation in different projects, where the quantification of
dynamical processes played a central role and needed to be addressed spatially distributed and
not only as singular point measurements. Thus, the need of remote-sensing based methods
arose and was applied. In Iceland, the main scientific problem of the project was related to
mapping and distribution modelling of permafrost. In earlier investigations in this area rock
glaciers were assumed not to be related to permafrost, and assumed to be relatively young
features from the Little Ice Age. Velocity and surface change investigations could contribute
3
to this discussion. On Svalbard, the coastal areas are active areas of erosion and contribute
with a sediment-input to the Arctic Ocean. The Arctic Coastal Dynamics project (Rachold et
al. 2005) aims at quantifying these processes in the circum-Arctic, however, mainly related to
soft sediment coasts. A major question for the Norwegian group was how to treat coastal
bedrock cliffs, which dominate on Svalbard, in terms of erosion. Thus, terrestrial
photogrammetry was used to test this question. The outlets glaciers of Jostedalsbreen are
highly dynamic features. During the last decade many outlet glaciers experienced great
advances, with a following increase of possible ice avalanche hazards. This is the background
for the glaciers being part of a European project aimed at assessing such hazards (Glaciorisk)
and the need for high-resolution surface velocity patterns arose.
1.2
Objectives
Within this thesis remote-sensing-based techniques are applied to measure and
quantify processes that govern geomorphic development in selected periglacial and glacial
environments. The main objective of the thesis is to address and validate these remote sensing
techniques for geomorphic process quantification in terms of applicability and accuracy. The
accuracy limits of the different techniques, and what constraints the accuracy sets for result
interpretation is assessed in the thesis. The sub-aims of the thesis are related to the
applicability of the techniques within the particular study areas.
Multi-temporal photogrammetry is used to quantify rock glacier and creeping slope
landform dynamics in Iceland. Little is known about the dynamics of these landforms in this
highly maritime climate, which actually have a central place in an ongoing debate on rock
glacier definitions, formations and relations. Thus, basic questions were related to the
dynamics in space (velocity fields) and time (travel times and landform age) derived from
photogrammetry, such as: Does the landforms move actively and at what rate? Has the
movement changed through time? What causes the movement, and how old are the
landforms?
Terrestrial photogrammetry is used to quantify the costal cliff retreat rate on Svalbard.
The main aim was related to estimating retreat rates of rock cliffs, which is a major objective
for quantifying arctic coastal dynamics in permafrost environments as a response to climate
change and sediment supply to the arctic basins. Until now, most costal cliff retreat
measurements have been carried out in soft sediments. Thus, can terrestrial photogrammetry
provide realistic and useful estimates for such quantification for coastal bedrock cliffs?
4
Multi-temporal photogrammetry and satellite SAR interferometry are applied to
quantify glacier movement. Also in this case spatially distributed surface velocity patterns are
important measures for glacial dynamics and as a validation tool for numerical glacial
modelling. Glaciers react much faster to climate variability than the other processes studied in
this thesis, so high-resolution data sets are often necessary. Main questions are related to the
applicability of these techniques in deriving high-resolution data sets, and their limitations.
Can cross-correlation matching of orthophotos be used for high-resolution mapping of glacier
surface velocities? How can we derive glacier velocities from height differentiated InSAR
scenes using non-interferometric conventional GIS software? How do the mapped velocity
fields influence surface elevations change of the same advancing glaciers outlets of
Jostedalsbreen?
The final aim of the thesis is the summary of some basic rules and recommendations
in relation of the applicability of various remote-sensing based techniques within quantitative
geomorphic process assessment.
1.3
Outline
The thesis is composed of two parts:
Part I provides an overview, where this introduction constitutes the first chapter.
Chapter 2 deals with geomorphic processes in periglacial and glacial environments and
chapter 3 with methods for measuring these processes focusing on methods using remotely
sensed data. Chapter 4 presents a summary of the research and main results from the papers.
A general discussion of these results is given in chapter 5. Conclusions and recommendations
for further research are found in chapter 6 and 7.
Part II contains the full version of the four papers comprising the main scientific
work of the thesis. The papers are referred to by their Roman numbers elsewhere in the text.
5
2 Geomorphic
processes
in
periglacial
and
glacial
environments
Ahnert (1998) identifies three research stages of both systematic and regional
geomorphology, namely morphography, functional geomorphology and historic–genetic
geomorphology. Morphography aims at describing landforms and their geometric
characteristics and dimensions, record relevant material properties and finally to identify
specific traces of processes. This information forms the basis of functional and historicgenetic geomorphology. Functional geomorphology aims at quantitatively expressing the
relationship between landforms, rocks and soil material and processes. The third and last
stage, historic-genetic geomorphology, deals with the formation and long term development
of landforms through time and changing climates.
Functional geomorphological systems have components dealing with form and form
characteristics, types of material and material characteristics, and processes and process
characteristics. Process characteristics include runoff frequency, erosion rates, debris velocity
etc.
This thesis is written within the scope of remote sensing applied within functional
geomorphology focusing on measuring geomorphic process characteristics using aerial and
satellite imagery.
2.1
Geomorphic processes
Geomorphic processes are either endogenic or exogenic. Endogenic processes
originate in the earth’s interior resulting in crustal movements or volcanic activity. Endogenic
processes increase the relief and thereby the potential for exogenic processes. Exogenic
processes originate outside the solid earth, in the hydrosphere, atmosphere or space. These
processes include weathering, denudation, fluvial processes, glacial processes, littoral
processes and aeolian processes. Exogenic processes are therefore highly climate dependant
(Ahnert 1998). Since measuring exogenic processes is the topic of this thesis, the following
will focus on these.
The rates and magnitudes of geomorphic processes vary with time. Some processes
are acting continuously, while others occur episodically. For episodic processes the recurrence
7
frequency is important for landform genesis; extreme and high-intensity events occur more
seldom than low- to medium events that dominate most of the time. This magnitudefrequency relationship is an important property of geomorphic processes in a given
environment and the may change with a changing climate (Summerfield 1991). Continuous
processes like glacier flow will off course also change with climatic factors.
To be able to compare the effectiveness of different geomorphic processes the concept
of geomorphic work was introduced by Caine (1976) based on earlier works of Rapp (1960b)
and Jackli (1957). Rapp (1960b) and Jackli (1957) use “exogenic mass transfer” to compare
different slope processes. This measure is tons of mass transported through a vertical distance
in unit time (i.e. ton-metre per year or cubic metre-metre per year). Caine (1976) argues that
erosion of a landscape is movement of material from higher to lower elevation, and therefore
a form of physical work. He therefore introduces the gravitational constant into the equation
of Rapp (1960b) and Jackli (1957) to convert mass transfer measured in ton-metre to work
measured in joules.
'E
mg (h1 h2 )
Where 'E is the energy (work) needed to transport the mass m through the vertical
distance h1 h2 , where g is the gravitational constant (9.8 m s-2). For a mass of volume v and
density U to be transported down a slope of distance d and slope angle
T , the energy needed
will (Caine 1976):
'E
vUg (d sin T )
Eq. 1
An erosion rate (rate of energy conversion through time) can then be defined as a
power and measured in watts. Comparing the work of the slope processes active in a small
mountain catchment in Colorado, Caine (1976) finds mudflows to be the most effective. In
comparison Rapp (1960b) finds solute loss to be the most effective process in the periglacial
area of Kärkvagge, northern Sweden having an relative importance of 48% and followed by
earth slides (34%).
The rates at which processes operate on a certain slope vary greatly through time. For
instance the large proportion of movement by earth slides in the study of Rapp (1960b) was
mainly achieved in one single storm event. As shown by Young and Saunders (1986) there is
also a considerable variation in the activity of a particular process under different
8
morphoclimatic regimes. This again emphasizes the climatic importance on geomorphic
processes and process rates.
2.2
Periglacial and glacial environments
The word periglacial literally means around the glacier and in this thesis the definition
of Washburn (1979) stating that “the term periglacial designates primarily terrestrial, nonglacial processes and features of cold climates characterized by intense frost action, regardless
of age or proximity to glaciers” is used. In the Multi-Language Glossary of Permafrost and
Related Ground-Ice Terms compiled by the International Permafrost Association's
Terminology Working Group (van Everdingen 1998) periglacial is defined as “the conditions,
processes and landforms associated with cold, non-glacial environments”. Both definitions
exclude the vast permafrost underlain boreal forests from the periglacial zone. According to
(Williams & Smith 1989) the periglacial environments cover 35% of the earth’s land area, and
mainly in the northern hemisphere. The periglacial zone roughly coincides with the areas
above and north of the timber line in the northern hemisphere, and above and south of the
timber line in the southern hemisphere. Glacial environments denote the areas covered by
glaciers and currently influenced by glacial processes. Glaciers cover 10% of the earth’s land
surface (Paterson 1994) with 84% and 12% of the area accounted for by the Antarctic and
Greenland ice sheets (Benn & Evans 1998). Both periglacial and glacial environments are
confined to high latitudes or high altitudes.
These areas has become of increased interest due to their high sensitivity to climate
change. In both environments seasonal and perennial frozen water plays a crucial role.
Changes in the state of water are governed by the energy balance at the ground surface, and
thus vulnerable to climate variations. In following the geomorphic processes of periglacial
and glacial environments dealt with in the four papers are explained further.
2.2.1 Slope movements and permafrost creep
The slope movements investigated in this thesis are all slow moving and situated
within the periglacial zone (Paper I). Landslides usually move along a sliding surface within
the rock /debris/soil body. The sliding surface could be concavely upward-curving (rotational
slides) or near-planar (translational slide). The sliding surface for rotational slides is often
9
found in alternating rocks and sediments which have variable permeability and strength
properties, and within bedrock with horizontal bedding planes (Cruden & Varnes 1996, Dikau
et al. 1996). For translational slides the movement is controlled by surfaces of weakness
within the body. The discontinuities could be intersecting joint surfaces, bedding planes,
faults, thrusts and depositions have varying shear strength and permeability (Cruden & Varnes
1996, Dikau et al. 1996). Slides may occur in one singular event or as several events separated
in time. Hence, the movement is episodic and may be triggered by a wide range of factors like
human intervention by construction work, climatic factors like heavy rainfall, earth quakes,
undercutting by river or wave action etc. Landslides may also develop into more fast moving
mudslides, debris flows etc. depending on the nature of the slide and material involved.
Landslides can display a wide range of velocities from some millimetres per year to metres
per second and appear in all climatic regions.
Given a cohesive material the whole mass may continuously deform due to
gravitational forces. Such cohesive flow may occur in permafrost areas. Permafrost is defined
as ground material with temperatures below 0ºC for more than two years (Williams & Smith
1989). In areas of permafrost unconsolidated material supersaturated with interstitial ice may
move downslope due to deformation ice. This is called permafrost creep and most often
expressed through the landform rock glacier (following the Barsch (1996) and Haeberli (1985,
2000) definition of a rock glacier). Rock glaciers can develop beneath talus slopes and within
glacial debris from end moraines by the incorporation of refrozen meltwater. If a slope is
present the material will start to creep. See Fig. 1 on rock glacier development. Deformation
of sedimentary glacier ice, leading to movement of debris covered glaciers or glacier-derived
rock glaciers (Humlum 1982) will be discussed in the next section
Permafrost creep has mainly been measured within rock glaciers and display velocities
from centimetres to metres per year (see Barsch for an overview (1996)). The velocity of rock
glaciers depends on ice content, applied shear stress, thickness, grain size and form,
temperature regime and ice density (Barsch 1996). The horizontal velocity pattern of rock
glaciers show increasing velocities from the rooting zone towards the front and the maximum
velocity is found close to the front where the rock glacier is at its thickest. Bottom topography
may alter this pattern and cause compressive flow with furrows and ridges close to the front
(Barsch 1996). Kääb & Weber (2004) have investigated the development of transverse ridges
on rock glaciers through both field measurements and laboratory experiments and conclude
that they are caused by compressive flow that may be induced by a break in the underlying
slope. Looking at a cross sectional profile the velocity increases towards the centre.
10
Fig. 1: Morphological model of rockglacier development from Haeberli (1985). Early
development shows a terrace-like form at the foot of perennially frozen talus. Large
rockglaciers may form from this stage. In some cases small glaciers may exist on top of the
rockglacier (Haeberli 1985).
Lichenometric studies have revealed that the surface does not show different shearing
movements and that the surface therefore is passively riding on a creeping frozen core
(Haeberli et al. 1979). Borehole deformations have revealed that a thin layer half way within
the rock glacier accommodates 75% of the deformation causing the horizontal movement at
the surface (Wagner 1992). When it comes to vertical velocities they are usually much smaller
than the horizontal and are related to loss of ice by melting (subsidence), movement along the
slope gradient and gain of mass by refreezing of meltwater. At the front of active rock glaciers
frozen debris over the front and partial melt out take place. The material is then incorporated
and subsequently overridden at the base. The advance mechanisms of rock glaciers are
11
therefore described as ‘conveyor belt’-like (Haeberli et al. 1998, Kääb & Reichmuth 2005).
The deformation of the ice within a rock glacier depends, among several factors, on the ice
temperature. And seasonal variation of surface velocity with higher velocity during summer
(Haeberli 1985) have been detected. Recently a more long-term speed up of rock glaciers
have also been seen in Turtmann valley, Switzerland and is explained by temperature rise in
the 1990s (Roer et al. 2005).
However, it is worth noting that there is some semantic controversy regarding the
definition of the term rock glacier and the different views may be explained by different
research approaches (i.e. morphography vs. functional process orientated geomorphology).
Barsch (1996) and Haeberli (1985, 2000) use the term rock glacier (in one word) solely for
perennially frozen debris supersaturated with interstitial ice and ice lenses that endure creep
due to deformation of ice. This includes the cohesive flow of talus rock glaciers and debris
rock glaciers, but excludes features originating from sedimentary glacier ice and therefore
emphasizes the process involved. Humlum (1982), having a morphographic approach,
distinguishes between talus- and glacier-derived rock glaciers (in two words), independent of
ice origin, but with the precondition of permafrost existence. Whalley includes even debris
covered glaciers in non-permafrost environments in his definition (Whalley 1974). In the
preceding sections the Barsch and Haeberli definition is used. Deformation of sedimentary
glacier ice will be discussed below.
According to Caine (1974) an alpine drainage basin can be divided into two
subsystems if geomorphic activity; the valley slopes and the stream channel of the valley
floor. The slope system involves processes of input by weathering, transfer by mass wasting
and storage by colluviation. The stream channel system takes over at the bottom of the slopes.
Considering slopes ending at sea level, as is the case for some of the study areas in this thesis,
the marine or littoral system would replace the stream channel system at the end of the slopes.
The input of material to the slope system is through weathering, which will be discussed later.
Mass wasting can take the form of several different processes in alpine environments and both
slow moving slides and rock glaciers are among these. An inactive slide or rock glacier can be
seen as storage. Barsch & Caine (1984) use a further subdivision of slopes in their
morphodynamic systems of high mountain areas, namely the glacial system, the coarse debris
system, the fine sediment system and the geochemical system. Using this classification rock
glaciers fall within the coarse debris system and landslides, depending on grain size, may fall
within the fine sediment system. The use of this morphodynamic systems allow us to see
12
individual processes or landforms that we investigate in relation with the whole range of
geomorphic processes that shape the earth surface in the environment being studied.
2.2.2 Glacier dynamics
Papers III and IV, and to some extent Paper I are dealing with measurement of
glacier flow. From a geomorphological perspective glaciers can be regarded as landforms
themselves (Bloom 1998), but more important is their role as sediment transport system and
erosional agents.
The surface velocity of a glacier is a function of the two components; plastic
deformation of glacier ice and sliding along the sole of the glacier. The deformation rate ( Hxy )
of glacier ice is given by Glens flow law (Paterson 1994) and increase with shear stress (W xy ):
Hxy
AW xzn
Where n is a constant usually around 3 and A depends on ice temperature, ice crystal
orientation and the purity of the ice.
W xy is a function of glacier thickness and slope. If one
assumes that ice only deforms due to shear stress in the xy-direction the stream lines of the
glacier will be parallel to the glacier surface (i.e. laminar flow). The difference between the
surface velocity, us and bottom velocity ub or the velocity u(z) at depth z (i.e. the deformation
component) is given by (Paterson 1994):
us u( z)
u s ub
2A
( Ug sin D ) n (h z ) n 1
n 1
2A
( Ug sin D ) n h n1
n 1
Where ȡ is the density of the ice, g is the gravitational constant, Į is the slope angle, h is ice
thickness and z the depth. From the equation it can be seen that the velocity decrease with
depth and that the greatest change in deformation and velocity occurs close to the bottom of
the glacier since the bed exerts a basal drag on the glacier. This is caused by the friction
between the glacier sole and the bed beneath the glacier. In the same manner mountain slopes
along valley glaciers will exert a drag and slow down the glacier and one therefore usually
introduce a shape factor in the equation for shear stress. The glacier thickness also decreases
toward the sides and this will decrease the velocity towards the edges. Hence, the velocity is
greatest along the centre of the glacier.
13
Ice movement or sliding along the glacier bed implies that the glacier is not frozen to
its bed. The temperature therefore has to be at the pressure melting point. Some glaciers (coldbased or polar glaciers) are frozen to the bed and plastic deformation will then be the sole
contributor to surface velocity. According to Weertman (1957) sliding can be explained by
two mechanisms; regelation and plastic deformation. Increased pressure on the upstream side
of bumps and melts the ice. On the downstream side of the bump the pressure is lower and the
water refreezes. This is regelation and will not occur on large bumps (order of 1m or more in
length). For larger bumps and for all ice below the pressure melting point increased pressure
on the upstream side of bumps will lead to increased plasticity in the ice. The velocity around
the bumps will therefore increase as well. The amount of water and water pressure at the
bottom is also important for glacier sliding. Iken & Bindschadler (1986) found a very good
correlation between measured sub glacial water pressure and surface velocity at
Findelengletcher.
To retain geometric equilibrium the net accumulation above any given point on the
glacier has to be transported beyond. This can be shown by the continuum equation. The flux
through a cross section of the glacier therefore has to be the same as the net accumulation
above the cross section. The velocity therefore has to increase down to the equilibrium line,
where the net balance is zero. Due to an increasing negative net mass balance the velocity will
decrease from the equilibrium line towards the front. In the accumulation area, where the
velocity generally increases, on therefore gets extending flow while there will be compressive
flow in the accumulation area. There will also be extending and compressive flow in areas of
changing bottom topography and in areas of widening and narrowing of the glacier. Generally
the velocity of a valley glacier will be greatest where the centreline meet the equilibrium line
and decreasing to either sides of the glacier and decreasing in the downstream and upstream
direction as well.
If the glacier it not underlain by bedrock but a layer of till deformation within this
subglacial till will also contribute to the surface velocity. Boulton & Jones (1979) showed that
near the margin of Breidarmerkurjökull, shear in water-saturated till accounted for 90 % of
the velocity measured at the surface.
Returning to Barsch & Caine (1984) morphodynamic systems of mountain areas,
glaciers can be seen as an independent system responsible for transport, erosion and
deposition. The glacial system is rather effective when it comes to transport to the glacier
margin where it interact with the fluvial system. Glaciers also has a dominating influence on
14
erosion in glaciated areas due to its place among the most effective agents of erosion on earth
(Benn & Evans 1998).
Several different models exist for describing subglacial friction, see Benn and Evans
(1998) for an overview. This friction has an impact on subglacial erosion, transport and
deposition. The three models apply for different basal debris concentrations. Both the
Coulomb friction model introduced by Boulton (1974, 1975, 1979) and the sandpaper friction
model (Schweizer & Iken 1992) includes the glacier thickness as on of its variables. Hallet’s
(1979, 1981) friction model does not include glacier thickness but ice velocity component
normal to the bed. The sliding velocity is also an important control on subglacial abrasion
rates (Benn & Evans 1998). This means that glacier thickness and sliding velocity is
important for the interactions between a glacier ant its bed.
2.2.3 Glacier surface change
Paper III is also dealing with glacier surface change. Due to their high sensitivity to
climate change glaciers are important for monitoring purposes. Changes in thickness and size
of glaciers will also affect their geomorphic effectiveness regarding material transport and
erosion. Hence, monitoring glacier surface change is also important from a geomorphic
perspective. Glacier surface elevation change is a combination of mass balance and flux
gradient (Paterson 1994):
wh
wt
b
wq
wx
Where h is glacier thickness at time t, q is the mass flux along the flow line x, and b is the net
mass balance. Changes in the density of snow and ice can also cause geometrical changes.
Dynamic properties will changes according to the parameters explained in the preceding
sections.
Mass balance is the difference between accumulation of snow and loss of mass
(ablation) through melting and calving. The energy used for melting can be broken down into
several parameters regarding long and short wave radiation, heat conductivity, latent heat,
surface albedo etc. (see Paterson (1994) for an overview). However, it has been shown
empirically that the summer temperature expressed through the sum of positive degree-days
can be used to estimate melt, due to the high correlation between the two quantities
(Braithwaite & Oleson 1989). For simplicity the mass balance can therefore be said to depend
15
on winter precipitation and summer temperature. A change in either two of these climatic
variables will therefore change the mass balance.
It is also important to bear in mind, that changes in the mass balance and thereby
geometry of glaciers will haven an effect on the dynamics. As seen changes in thickness will
have an impact on the velocity. Changes in mass balance will also lead to change in the
frontal position of glaciers. Response time is the time needed for a glacier to adjust to change
in mass balance and regain equilibrium. Reaction time is the time interval from a change in
mass balance occurs until frontal changes can be seen. Both response and reaction time vary
greatly with ice thickness, mass balance characteristics, steepness and length of the glacier,
but the reaction time is shorter than the response time. As a result of changes in summer
temperature and winter precipitation most glaciers in the world have retreated during the last
decades (IAHS(ICSI) 1998, IAHS(ICSI) 2001). As a widely recognized exception from this
global trend the glaciers in the western part of Norway have been advancing from the late
1980s up until the start of this millennium. The surface changes of some of the glaciers in
western Norway are investigated in Paper III in this thesis. The advancement has been
related to increased winter precipitation in the late 1980s and beginning of the 1990s
(Kjøllmoen 2003, 2004).
2.2.4 Cryogenic weathering and permafrost coastal cliff retreat
Cryogenic weathering is the combination of mechanical and chemical breakdown of
rock in cold climates. Mechanical breakdown of rock by freezing of water is pore spaces,
joints and bedding planes is taken to be a very effective geomorphic process in periglacial
environments resulting in angular rock fragments. Earlier the general belief was that the
number of freeze-thaw cycles was the most important for frost shattering of rocks. But in the
last decades it has been shown that stable temperatures conditions below 0°C are the most
effective. According to Hallet et al. (1991) micro fraction propagation happened at
temperatures between -3°C and -6°C. From field studies in Japan, Svalbard and Antarctica it
has also been found that the frost shattering rate depends on degree of water saturation and
tensile strength of the rocks, and that moisture content is the most important factor controlling
the shattering rate (Matsuoka 1990, 1991). Lately the almost neglected effect of chemical
weathering has been brought to attention (Hall et al. 2002).
When it comes to periglacial and permafrost coasts, frost weathering was already
proposed by Høgbom (1914) and Nansen (1922) to be an important process regarding the
16
formation of the coastal cliffs characteristic of arctic fjords. Nansen (1922) also suggested that
the strandflat forms because of the interaction of frost weathering and marine processes at the
base of the cliffs, with frost weathering as the main fracturing agent and marine processes
removing the weathered material. The appearance of the coastal rock walls together with the
amount of angular rock fragments accumulating on the snow and ice-foot below coastal cliffs
during spring show that subaerial weathering is active and important. Arctic coasts are in
general low energy coasts since sea ice and pack ice shortens the fetch or completely obstructs
the possibility of wave action for several months of the year (French 1996). In some areas the
snow and ice could survive all the way through summer above the high tide, indicating that
the removal of material is restricted to heavy storms recurring in the fjords at several years
interval (Ødegård & Sollid 1993). It has been observed that this ice-foot plays an important
role in both removal of weathered material from the base of arctic coastal cliffs and fracturing
the lower part of cliff walls (Nielsen 1979). It has also been shown that the temperature
regime of coastal bed rock cliffs in permafrost areas are favourable for cryogenic weathering
processes with stable sub-zero temperatures and steep temperature gradients throughout the
spring (Ødegård & Sollid 1993, Ødegård et al. 1995). The presence of snow and ice-foot in
the same period will also provide the sufficient moisture.
17
3 Remote sensing of geomorphic processes in periglacial
and glacial environments
Remote sensing is defined as the gathering of information at a distance (Campbell 1996)
and comprises the use of several different techniques and platforms. Some authors also define
remote sensing as the whole system of acquiring, processing and output. The use of aerial and
terrestrial photogrammetry is some of these remote sensing techniques and developed from
the start of the last century mainly as a tool for conventional mapping. Satelliteborne synthetic
aperture radars (SAR) were first launched in the late 1970s and the use of satellite SAR
interferometry emerged in the early 1990s. The following sections give the background for
the remote sensing techniques used in this thesis and compare them with other techniques and
discuss the applicability of the techniques for measuring geomorphic processes in periglacial
and glacial environments.
3.1
Digital photogrammetry and optical remote sensing
Laussedat compiled topographic maps from terrestrial photographs in 1849 and is
regarded as the father of photogrammetry. He also created the first camera suitable for
photogrammetric measurements together with a procedure called metrophotographic. This
was preceded by the production of the first usable photographs presented by Niepce and
Daguerre in 1839. The word photogrammetry was first used in a paper by Meydenbauer in
1893. Wilbur Wright was the first to take photographs from an air craft in 1909. The first
stereoautograph was produced by von Orel and Zeiss for plotting of terrestrial photographs in
1911. Starting as plane table photogrammetry in the 1850s, and continuing as analog
photogrammetry from around 1900, photogrammetry went into the analytical stage with the
development of the analytical plotter by Helava around 1960. Analytical photogrammetry
improved for some decades with the progress in electronic computation and during the last
two decades the procedure of photogrammetry has turned 100% digital with the use of digital
photogrammetric workstations (DPWs). (Section based on Mikhail et al. (2001) and Burtch
(1997)).
19
3.1.1 Geometry
Assuming the geometry of Fig. 2 the elevation, h, of point A, can be determined based
on the parallax of the same point in the two aerial photographs (Mikhail et al. 2001).
H h =
Bf
p
Where B is the basis between the two projection centres, L1 and L2, H is the flying height
above sea level, f is the focal length of the camera used and p is the x-parallax of point A in
the two photographs. More often parallax is measured at two close points and the difference
in parallax used to deduct the elevation difference, ǻh (Mikhail et al. 2001).
'h h2 h1
§ 'p ·
¸¸
Bf ¨¨
© p1 p 2 ¹
Where p1 and p 2 is the parallax of point 1 and 2 with elevations of h1 and h2 , and the
parallax difference 'p = p 2 - p1 .
Fig. 2: Principal geometry of stereo photogrammetry from Mikhail et al. (2001) showing
parallax p at a point A. O1 and O2 are the principal points of the two photographs with
projection centres L1 and L2. The parallax p is xl - xr , where A is imaged as xl and xr in the two
photographs.
The ideal geometry of Fig.2 is usually not the case. One therefore has to determine the
position (Xc,Yc,Zc) and the rotation elements (Ȧ, ij, ț) of each camera station L1 and L2 (i.e.
exterior orientation). These parameters can be obtained using GPS and INS (Inertial
20
Navigation System) aboard the air plane when acquiring the air photos or through relative and
absolute orientation of the photogrammetric model, given that interior orientation already is
performed. The scope of interior orientation is to establish the transformation between the
pixel coordinate system of the digital image and the camera coordinate system. Using
photographs from digital sensor this is done automatically. Using scanned air photos it is done
by locating fiducials with known coordinates in the camera coordinate system in the digital
image. Information from camera calibration reports on the focal length of the camera, location
of principal point and lens distortion should also be supplied. Relative orientation can then be
done in order to solve 5 of the 12 unknown parameters of exterior orientation. In digital
photogrammetry dependent orientation is used for relative orientation and done by measuring
five or more corresponding points in the two photos. The rotation and translation elements of
one photograph with respect to the other are then solved. The result is a stereomodel.
Absolute orientation is done by applying at least three ground control points with known
elevation and at least two points with known x and y coordinates, to solve the remaining
unknown parameters of exterior orientation. The three vertical control points level the model
through solving the parameters for rotation around the X and Y axis and solving the Ztranslation (in ground coordinates). The two horizontal control points solve the translation
along the X and Y axes and the rotation around the Z axis, in addition to determining the scale
of the model. By absolute orientation the transformation between the stereo model generated
during the relative orientation and a known ground coordinate system is established (i.e. the
position (Xc,Yc,Zc) and the rotation elements (Ȧ, ij, ț) for each camera station L1 and L2 are
determined). The use of additional points for both relative and absolute orientation allows
errors to be estimated and will also increase the accuracy of the photogrammetric model.
After the steps of orientation had been performed the photogrammetric model has a uniform
and known scale and three dimensional ground coordinates for every location in the model
can be extracted. This means that mapping can be done and DTMs produced. Positional
accuracy of 0.3 pixels in the XY-plane and 0.01 - 0.06 % of the flying height in Z can be
obtained within a DPW. These accuracies are the same as the ones achieved by analytical
photogrammetry (Maalen Johansen & Andersen 1998).
3.1.2 Automatic generation of DTMs
Having an exterior oriented photogrammetric model, DTMs can be generated
automatically within a DPW. The methods are base on locating homologous points in the
21
photos that make up the stereomodel and measure the x-parallax of these points. As explained
earlier the x-parallax is an expression of elevation or elevational differences. The user
specifies a resolution of the DTM and an elevation is calculated for each grid cell of the DTM
based on the stereomodel. This is done by stereo image matching, where homologous points
are located by the use of some kind of automatic matching algorithm. Image matching
algorithms are classified into two approaches: area matching and feature matching. Feature
matching is not commonly used for this purpose, but is based on detecting interest points and
line segments in the two photos and matching them. In area matching the grey levels in a subwindow of one photo (reference window) is matched against the grey levels in a greater subwindow (test window) of the other photo. Area matching is typically performed by calculating
the normalized correlation between reference and test windows. The normalized correlation,
J , between pixel values of the two windows is calculated as the covariance of the two
windows divided by product of standard deviation of each window (Gonzalez & Woods
1993):
J s , t ¦ ¦ > f ( x, y ) @ >w( x s, y t ) w@
¦ > f ( x, y ) f ( x, y )@ ¦ ¦ >w( x s, y t ) w@
x
¦
x
f ( x, y )
y
2
y
2
x
Eq. 2
y
Where f ( x, y ) is the test window of size MxN, w( x, y ) the reference window, s, t the
position of the reference window with s  (0, M 1) and t  (0, N 1) , w is the average
value of w( x, y ) and f ( x, y ) the average of f ( x, y ) in the region coincident with the
current location of w . When covariance is calculated the average value of each window is
subtracted (see Eq. 2). This operation in addition to the division by the standard deviation (i.e.
normalization) accounts for differences in brightness and exposure between the two photos
and improves the result of the method considerably. The normalized correlation will be a
value in the range -1 to 1 and is calculated for each possible position of the reference window
within the larger test window, and the location that yields the greatest normalized correlation
value is taken to be the position of the reference window within the test window. The same
procedure is sometimes referred to as the correlation coefficient or the double crosscorrelation function.
22
The x-parallax is then measured based on the position of the reference window
location in both photos and converted into an elevation value. One cell within a DTM is
usually based on several matching measurements. If the correlation value does not reach a
prescribed threshold the elevation value in the resulting DTM is either marked in some way or
not calculated at all. Similar methods are used for different tasks within a DPW (e.g. locating
fiducial marks during interior orientation, automatic relative orientation and refinement of
control point measurement etc.) and also for calculation of displacement within multitemporal orthophotos or satellite scenes (treated later). Correlation functions are also used for
different purposes within general digital image processing. Some matching algorithms use
image pyramids, meaning that preliminary matching is done on courser versions of the air
photos and refinement of the matching result is performed with photos of the successively
better resolution. Some matching routines are designed to work in the frequency domain and
the algorithm then includes some kind of Fourier-transform.
Before automatic generation of DTMs are done some DPW also requires resampling
of the air photos into epipolar geometry. An epipolar plane is the plane made up of the camera
centres (i.e. projection centres) of the two air photos in the stereo model, a point on the
ground and the corresponding image points in the two air photos (In Fig. 2 the epipolar plane
is the triangle defined by L1,L2 and A). The epipolar line is the line between the image points
(al and ar in Fig. 2). When the stereomodel is fully orientated it is possible to transform the air
photos into epipolar geometry. The epipolar lines will represent the lines in the new image.
Having images with epipolar geometry facilitates stereo viewing but it also reduce matching
of corresponding points to one dimension, since an object on the ground will be pictured
along the same epipolar line in the two images of the stereomodel.
According to Fryer et al. (1994) the accuracy of elevation measurements done by
aerial photogrammetry is about 1-3 parts per 10,000 of the camera/object distance rising to 10
-30 parts per 10,000 with sparse ground control. Their study was done using analog
photogrammetry, but Baily et al. (2003) have shown that the accuracy of digital
photogrammetry for geomorphological purposes should be the same. For widely used flying
heights this gives accuracies in the range of decimetres to metres. The accuracy also varies
with topography with greater precision of gently sloping terrain than for more mountainous
and ragged reliefs.
Optical satellite imagery can also be used for generation DTMs. The imagery can
either be from different satellite tracks or from simultaneous acquisition from two sensors on
the same platform (one nadir- and a forward- or backward–looking sensor). Multitemporal
23
SPOT data (10 m resolution) has been used for creating DTMs in mountainous regions (AlRousan & Petrie 1998, Bishop et al. 2001, Zomer et al. 2002). DTMs have also been
generated with sensor as Ikonos (1 m resolution), Aster (15 m resolution) and Landsat 7 (15 m
resolution) (Toutin 2001, 2002). ASTER on the Terra satellite has a backward looking sensor
in addition to the nadir looking and are capable of creating along-track stereo. The techniques
for DTM generation from satellite stereo data are similar to aerial photogrammetry, but earth
rotation, earth curvature etc. have more effect. On the other hand, satellites are more stable
platforms and less disturbed by rotation etc. Eckert et al. (2005) found that given accurate and
well-distributed ground control points (GCPs), it is possible to generate DTMs with a root
mean square (RMS) error between 15m and 20m in hilly terrain and about 30m in
mountainous terrain from ASTER data.
3.1.3 Orthophotos
Orthophotos are air photos transformed into orthographic projection. Original aerial frame
imagery is based on light passing through a single point, the perspective centre, before hitting
the image plane. Points at the same horizontal location but with different elevation will be
pictured at different locations in the image (i.e. relief displacement). The scale of the original
air photo also varies with elevation, due to the central projection. Using an orthographic
projection the light rays are perpendicular to the chosen horizontal reference plane (i.e. datum
plane). Changing elevations will not have an effect in this projection and the scale throughout
the orthophoto is also constant. To produce an orthophoto a DTM is needed in addition to the
exterior orientation parameters of the original air photo. Using this information the original air
photo can then be reprojected in to an orthophoto.
There are two basic approaches for generating orthophotos; forward projection and
backward projection. Using forward projection pixels from the original photo are projected
onto the DTM to determine their spatial map coordinates and these space points are then
projected into the orthophoto. The space between the points projected into the orthophoto
varies from terrain variation and perspective effects, the final orthophoto pixels are
determined by interpolation. Using backward projection the X and Y map coordinates for the
final orthophoto is first calculated. Then the elevation at each of those X,Y location is
determined from the DTM and this spatial coordinate is used for retrieving a grey level from
the original air photo for the orthophoto. Interpolation or resampling has to be done in order to
determine the pixel value to be used from the original air photo. (Section based on Mikhail et
24
al. (2001). Z/I Imaging ImageStation, the DPW used for the work presented in this thesis,
applies backward projection method for creating orthophotos.
Since the reprojection to orthographic projection involves the use of a DTM, errors in
the DTM will lead to errors in the orthophoto. Since orthophotos are planar the error will
appear as horizontal displacements. Given a DTM error 'H and a focal length f , a point
located a distance r from the centre of the air photo measured in image coordinates will be
displaced 'R in ground coordinates.
'R
r
'H
f
For points located close to the corner of a 23 cm frame photo with a 0.15 m focal length a 1m
error in the DTM will lead to a 1.1 m displacement in X,Y-coordinates. The same error will
lead to a half the displacement for a point located half way between the image centre and a
corner. The resolution of an orthophoto is determined by the resolution of the original air
photos and the same resolution might be used for the orthophotos. The scale will therefore
depend on flying height, focal length of the camera and in the case of scanned analog
imagery, also the scanner resolution.
3.1.4 Mapping earth surface changes by DTM differencing
Difference between DTMs generated from multitemporal data may reveal changes in
earth surface elevation. Long-term mass balance of glaciers may be reconstructed using
DTMs constructed from series of multi-temporal air photos (Etzelmüller & Sollid 1996). The
difference is found by subtracting two DTMs. In order to minimize errors, known stable parts
of the terrain may be used for correction (Kääb 2004). The horizontal resolutions of
investigations by DTM differencing may be from centimetres for the case of close up
terrestrial photogrammetry, via decimetres for air borne sensors to 30 m for satellite borne
sensors like Aster. The vertical accuracy of DTM differencing depends on the accuracy of
both data set and can be written as the square root of the Pythagorean sum of the individual
accuracies of the two DTMs involved (Etzelmüller 2000):
etot
2
2
edtm
edtm
1
2
( Eq. 3)
25
This also means that DTM differencing is more strongly affected by noise than single DTMs,
and it is therefore common to apply some sort of smoothing filter on the DTM differenced
results. Measuring surface elevation change is very useful in geomorphology since many
geomorphic processes like erosion, advection and deposition result in surface elevation
change. The rate and spatial distribution of these processes can therefore be mapped by
differencing multitemporal DTMs. The method of DTM differencing is not limited to using
photogrammetric derived DTM from air photos or optical satellite images, but all methods
capable of generating DTM can off course also be used for DTM differencing (e.g.
tachymetry, laser scanning, interferometric SAR, radar altimeter etc.). DTMs have been in use
for the last decades and before this surface elevation change analysis was restricted to single
point measurements.
3.1.5 Mapping earth surface displacements from multitemporal optical
imagery
The idea of using multitemporal photographs for measuring earth surface movements
was conceived in the 1930s by Finsterwalder (1931) for measuring of glacier motion.
Terrestrial photographs were taken from the same position at some time interval. When two
such multitemporal photographs are analyzed in a stereo instrument the stable parts of the
terrain would not cause any stereo effect, while the parts influence by glacier motion would.
This stereo effect is not caused by parallaxes due to elevation but deformation. The
deformation parallaxes can be measured and transformed into glacier velocities in the same
manner as x-parallaxes from stereo photographs can be used for measuring elevation (see
section 3.1.1). The time interval applied must depend on the glacier velocity. The method of
Finsterwalder has also been applied to some of the glaciers used in this study. Pillewizer
(1950) measured the velocity of Nigardsbreen (Paper III) in 1937 and 1938, while Liestøl
used the method for the same glacier on several occasions from 1949 to 1961 (Østrem et al.
1976). Melvold (1992) used the method for Kronebreen on Svalbard (lower part of the glacier
investigated in Paper IV).
The method has later been applied to aerial photographs for
measuring the displacement of glaciers and rock glaciers using analogue and analytical
photogrammetry (Armenakis 1984, Brecher 1986, Haeberli & W. 1988, Gorbunov & Titkov
1992, Kääb et al. 1997, Kääb & Funk 1999).
By the advent of satellite remote sensing a similar and non-stereoscopic method is
used for measuring glaciers velocities. Recognizable features in two or more co-registered and
georeferenced satellite images could be used for manual measurement of glacier movement.
26
Lucchitta & Ferguson (1986) and Orheim & Lucchitta (1987) calculate glacier velocities in
Antarctica using 30 m resolution Landsat TM images using a method of manual feature
tracking of crevasses, rifts, depressions and tidal cracks. Feature tracking can also be done
automatically using cross-correlation algorithms. This was first demonstrated by Bindschadler
& Scambos (1991) also using Landsat TM images of Antarctica. Scambos et al. (1992) use a
similar algorithm named IMCORR which is a fast Fourier transform based version of the
normalized cross-covariance method. These methods were originally developed for accurately
locating tie-point pairs in two images for co-registration (Bernstein 1983) and are variants of
the normalized correlation method used for matching and automatic generation of DTMs
within a DPW (see section 3.1.2). Later similar studies of glaciers using SPOT and ASTER
data have been undertaken (e.g. Lefauconnier et al. 1994, Rolstad et al. 1997, Kääb 2002,
Kääb et al. 2002, Kääb 2005, Kääb et al. in press). In addition the new sub-meter resolution
satellite Quickbird show promising results using the same method for landslide deformation
mapping (Delacourt et al. 2004). Feature tracking is not limited to optical remote sensing, and
has also been applied to satellite SAR scenes. The technique is called speckle tracking and is
based on tracking point of high backscatter (speckle) in several multi-temporal SAR-scenes.
The resolution is limited to the spatial resolution of the SAR scenes usually around 20 metres
(Fahnestock et al. 1993).
The introduction of digital photogrammetry made orthophoto production easier, and
the same methods used for satellite images could therefore be used for orthorectified air
photos. Rolstad (1995) demonstrated this for velocity mapping of Engabreen, northern
Norway. Later cross-correlation matching of orthophotos has been widely used for measuring
movements of glaciers, rock glaciers and landslides (Baltsavias 1996, Kääb & Vollmer 2000,
Kaufmann & Ladstädter 2003, Delacourt et al. 2004).
According to Kääb and Vollmer (2000) the accuracy of this method is in the order of
one pixel and at least at the same level as results from conventional photogrammetry.
Meaning that displacements in the order of decimetres can be measured depending original air
photo resolution from flying height, focal length of the camera used and canner resolution. As
was the case for the method introduced by Finsterwalder on therefore has to adjust the time
interval between the acquisitions depending on the rate of the deformation in question.
Although originally developed for deformation mapping of rock glaciers using aerial
orthophotos, this method has also been used to map glacier movement using ASTER satellite
data (Kääb 2002, 2005, Kääb et al. in press).
27
3.2
Interferometric SAR (InSAR)
Radar remote sensing started with the launch of SEASAT in 1978. Since then
satellites like Radarsat, JERS-1 and Magellan have been in successful use. Interferometry was
first used for observations of the planet Venus and the Moon (Rogers & Ingalls 1969) and
later for topographical mapping of the earth surface from an airborne platform (Graham
1974). Goldstein and Zebker (1987) were the first to measure displacements of the earth
surface with two simultaneous operating airborne antennas and the first topographical maps
from spaceborne platforms where made from SEASAT and SIR-B (space shuttle) data
(Gabriel & Goldstein 1988, Goldstein et al. 1988) . Terrain surface deformation was first
mapped by Massonnet et al.(1993) investigating the displacements of Landers earthquake,
California. See Tab. 1 for the most frequently used satellites for interferometric purposes.
Ordinary Synthetic Aperture Radar (SAR) scenes give a two dimensional image of the
backscattered energy when transmitting a radar signal towards the Earth surface. These
images only show the amplitude (i.e. the power) of the received and processed signal.
Interferometric SAR (InSAR), on the other hand, uses the phase information in the radar
acquisitions. The technique uses two receiving antennas to produce an interferogram (InSARscene). The two antennas can be separated in either time (repeat track acquisition) or space
(along- or across-track). An interferogram is a 2-dimensional representation of the measured
radar phase difference between the two antenna positions and corresponding positions on the
Earth surface. This phase difference is made up of several factors where the most important
are: satellite viewing geometry, ground surface topography, ground movements between
acquisitions and atmospheric disturbance. Having accurate knowledge of the satellite orbit
and radar viewing geometry, one is able to extract the geometry factor from the phase
difference. Furthermore, correcting for the atmospheric distortion or assuming it to be very
small, one is able to use the InSAR technique to gather three-dimensional information on
topography and change of the Earth surface. For deformation mapping, earth surface
movements are usually far too small to be detected by single-track acquisitions (i.e. where the
two SAR antennas are mounted on an airplane). Instead, repeat-track acquisitions with
satellite SAR sensors can be used. To utilise this technique, the satellite has to pass in more or
less the same orbit twice.
28
Tab. 1: Characteristics for some of the most used radar satellites for interferometric purposes.
(† is out of service date).
Satellite
Launch date
Wavelength
Spatial
resolution
ERS-1
1991 (†2000)
56.7 mm
30 m
ERS-2
1995
56.7 mm
30 m
JERS-1
1992 (†1998)
235 mm
18 m
RADARSAT
1995
57.7 mm
8.4 m
ENVISAT
2002
56.7 mm
30 m
3.2.1 Geometry
The geometry of a satellite InSAR acquisition is shown in Fig. 3. The first SAR scene
is acquired in orbit O1 and the second in O2. These two orbits are separated by the spatial
baseline B, which can be decomposed into Bx and By. The distance from the two antenna
positions to a given point on the ground is r1 and r2, respectively. The difference in distance
from the point on the ground to the two antennas (r2 - r1) can be expressed as a certain number
of wavelengths and a fraction of a wavelength: a phase. This is the phase difference depicted
in the InSAR scene and can be written as (Gens & van Genderen 1996):
'I
4S
O
(r2 r1 )
4S
O
( Bx sin T B y cosT )
( Eq. 4)
Here, T is the look angle and O the wavelength, see also Fig. 3.
3.2.2 Unwrapping
Every pixel in the interferogram represents a phase difference. The phase differences
will vary in a continuous fashion for many surfaces and the phase differences in the
interferogram will also vary in the same way. The phase difference is measured modulo 2S; in
intervals from 0 to 2S. The phase differences in the interferograms are often displayed as
colour coded cycles or just grey level values in the range 0 to 255. It is these shifts from 2S to
0, or 255 to 0 speaking of grey values, that create the characteristic fringe pattern. Converted
to wavelengths, one fringe equals half a wavelength due to the double pathway.
29
Fig. 3: The geometry (side view) of an InSAR-acquisition. First acquisition for the InSARscene in O1 and second in O2. B is the spatial baseline between the two satellite orbits.
.
The fringes represent ambiguities and should therefore be unwrapped. The principal
method of all phase unwrapping is that the phase differences are integrated by adding
multiples of 2S when a fringe transition is crossed (i.e. when the phase difference wraps from
2S to 0). For a more thorough description of the different unwrapping methods and algorithms
see Gens & van Genderen (1996). After unwrapping, it is possible to create a DTM by using
one known elevation point together with a given spatial baseline to deduct the elevations for
every pixel in the whole scene.
30
3.2.3 DTMs from InSAR
The first application of SAR interferometry was topographic mapping using an
airborne system (Graham 1974). For air borne topographic mapping the two scenes needed to
make the interferogram are acquired simultaneously and the elevations accuracy is better than
2m. If slopes are to steep layover or shadowing may occur leading to ambiguous or useless
phase information. For satelliteborne systems like ERS-1 DTMs with relative errors of 5 m or
less have been derived (Zebker et al. 1994). This means that one should be able to detect
surface elevations changes in the order of 3 m for airborne systems and 7 m for satelliteborne
systems by differencing multi-temporal DTMs. It is worth noting that the microwaves
penetrate will penetrate materials like snow and ice. The ERS-1/ERS-2 satellites emit
microwaves with a 5.3 GHz frequency and they can penetrate 10 m of pure ice and 20 – 30 m
of dry snow (Ulaby et al. 1982). The penetration depth is highly dependant of the water
content and may therefore change between two acquisitions. Hoen & Zebker (2000) found
penetrations of 12 to 35 for ERS data from the Greenland Ice Sheet. Unless one has detailed
knowledge about these factors it is therefore difficult to perform multi-temporal differencing
of InSAR derived DTMs in glaciated areas. Terrain slopes and aspect can be derived directly
from interferograms without performing unwrapping (Wegmüller et al. 1993) and is very
useful for geomorphological studies.
Two antennas mounted on a space shuttle in February 2000 produced DTM from
interferometry during the Shuttle Radar Topography Misssion (SRTM). The mission
produced 30 m and 90 m resolution DTM for the entire globe between 60°N and 54°S. The
absolute vertical accuracy is 16 m and the relative vertical accuracy 6 m (Sun et al. 2003).
Rignot et al. (2003) have used SRTM data to document accelerated thinning of the Patagonia
Icefields in South America. Strozzi et al. (2003) compared SRTM data with an automatic
aerophotogrammetric DTM and found average height differences of 7.2 m.
3.2.4 Differential interferometry (DInSAR)
Satellite InSAR scenes are usually based on two scenes acquired some time apart. This
time interval is called the temporal baseline, in contrary to the spatial baseline mentioned
earlier. If a surface deformation has taken place during this time interval it will also cause
phase differences. The phase difference as noted in Eq. 4, could therefore be expanded with a
second part for the deformation (Kwok & Fahnestock 1996):
31
'I
4S
O
( Bx sin T B y cosT ) 4S
O
'U
( Eq. 5)
Where 'ȡ is the deformation component in range during the temporal baseline. Range is the
direction in which the radar signal is transmitted. Range is perpendicular to the flight
direction, which is also called azimuth. As one can see from Eq. 5 the contribution from the
deformation is only dependant on the temporal baseline and not the spatial baseline. In an
interferogram where deformation has taken place one has to isolate one of the two
components to utilise the other. To detect or measure deformation one has to remove the
contribution from the topography and vice versa. For deformation mapping it would have
been an advantage if the satellites went in exactly the same orbit for the two acquisitions.
Since this is not possible one need to remove the contribution of the topography. This is called
differential interferometry or DInSAR and can be done by using two interferograms (three
SAR-scenes) or one interferogram and a DTM.
Using two interferograms one has to assume that the motion is constant over the time
interval when all three scenes are acquired. One interferogram is made from the two first
scenes and one from the two last ones. If these two interferograms are subtracted the resulting
one will only show the topography. This again can be used to remove the topographic
contribution from one of the two original interferograms (assuming the atmospheric
contribution to be insignificant), and the result is a motion only interferogram. See Kwok &
Fahnestock (1996) for further details.
If there is already a DTM available, this could be used to remove the topographic
effect. The normal procedure is to simulate a SAR-scene by using the DTM to produce an
artificial interferogram that contains phase differences only due to topography. This artificial
interferogram is then subtracted from the original, resulting in a height-differentiated motion
only interferogram.
3.2.5 Limitations
The use of interferometry also has some limitations. One only gets a suitable
interferogram if the two signals received from the same object are coherent (i.e. not
decorrelated). This means that the phase of the signal has to be stable between the two
acquisitions. Decorrelation of the spatial baseline is a factor that is governed by the
32
acquisition parameters, but it can be avoided by choosing image pairs with a proper spatial
baseline. A longer spatial baseline will result in a smaller distance between the topography
induced fringes and at a certain spatial baseline the fringe pattern will be so dense that it
makes unwrapping impossible. This is called the critical spatial baseline and for ERS-1 and
ERS-2 it is about 1.1 km.
As noted before, the contribution of deformation to the phase differences is not
dependant of the spatial baseline. Another effect of the spatial baseline length is that shorter
baselines will result in less contribution from the topography. There are examples of applying
interferograms with short spatial baselines of only a few meters or for nearly flat ice sheet
surfaces, and thereby completely disregarding the topographic effect (Goldstein et al. 1993,
Michel & Rignot 1999). Temporal effects are changes that occur at the earth surface during
the time between the acquisitions and will also have an impact on the coherence of the
interferogram. This can be changes in the reflective properties of the material at the surface.
Changes in the water content will for example drastically change the dielectric properties of
the ground and, hence, also the backscatter of the electromagnetic waves. Motion at the Earth
surface could also decorrelate the image if it is too chaotic and has a discrete rather than
continuous fashion. Glaciers moving with high velocities (several meters per day) and with a
lot of crevasses will often decorrelate (Weydahl 2001).
3.2.6 Mapping earth surface changes and displacements by satellite
InSAR
After Goldstein et al. (1993) demonstrated the possibilities of satellite SAR
interferometry for mapping glacier velocity of an Antarctic ice stream, the technique has been
widely used for mapping glacier movement on for instance Greenland, Svalbard and in the
Alps (Joughin et al. 1995, Eldhuset et al. 1996, Rott & Siegel 1996, Unwin & Wingham
1997). All these have only measured the component of the glacial motion in the direction
towards the satellite, the so-called line of sight component. Traditional ground measurements
of glacier velocity are done in the direction of movement. Rignot et al. (1995) used existing
altimetric DTM to compute slope gradient and slope direction in Western Greenland. They
also made the normal assumption of glacier movement being surface parallel and calculated
velocities that were within 6% of ground measured ones. Kwok & Fahnestock (1996) used
two interferograms to separate motion and topography. The resulting DTM was then used to
33
decompose the glacier movement into the direction of slope. Longitudinal glacier strain rates
can be calculated without unwrapping of the interferogram since they depend solely on
changes in glacier velocities and no absolute velocity values are needed (Forster et al. 2003).
When measuring movement with InSAR data one only measures the component of
motion in the range direction. If all motion is taking place in the azimuth direction, there will
be no component in range and, hence, no motion induced contribution to the interferogram.
Mohr et al. (1998) combined interferograms from ascending and descending orbits to map an
ice stream on Greenland in three dimensions, and thereby omitting the problem of measuring
motion in the azimuth direction. They also assumed surface parallel flow.
The velocity field of unstable slopes and rock glaciers in periglacial areas have also
been mapped using InSAR. Rott et al. (1999) mapped a slope movement in the order of
millimetres to centimetres per year using ERS-1 and ERS-2 data from 1992 and 1998. Rignot
et al. (2002) measured velocities for rock glaciers in McMurdo Dry Valleys, East Antarctica
of up to 40 mm per year using the ERS-1 and ERS-2 satellites as well. This is a glacierderived rock glacier following the Humlum definition. Kenyi & Kaufmann (2003) found an
average deformation rate of 6 mm during 35 days and maximum deformation rates of 18 mm
in 35 days for a rock glacier in the Austrian Alps also using ERS-1/ERS-2 data. From the
results they also infer that the smooth appearance of the motion field supports the idea of ice
as a stress-transferring medium in rock glaciers. Using JERS-1 data in addition to ERS1/ERS-2 Strozzi et al. (2004) found interferometric methods to be suitable for area-wide
detection of rock glaciers and in individual cases also suitable for finding the order of surface
displacements. They also compare the results with photogrammetric results from aerial
photography and get the same results. In addition to the wide-range capability of DInSAR it
also detects movements of inactive or relict rock glaciers smaller than 5cm per year that are
not detectable with photogrammetry. Other processes like frost heaving have also been
measured using InSAR. Investigating an area at the foothills of Brooks Range in Alaska
Wang & Li (1999) found differential uplift of 3 centimetres caused by frost heave during from
August 30 to October 4 1995 using ERS-1 and ERS-2 data and state that the technique could
be used for regional monitoring of changes of the active layer.
When it comes to the accuracy of satellite interferometry it is some fractions of the
wavelength used and varies from some millimetres to centimetres. Goldstein et al. (1993)
state that the detection limit for 5.6 cm wavelength ERS-1 data is about 1.5 mm for vertical
motions and about 4 mm in for horizontal motions. Massonnet and Feigl (1998) present errors
34
of 5mm, 6mm and 37 mm from interferometric studies of three different earthquakes. The
propagation of topographic errors contributes with 4mm, 3mm and 9mm respectively. The
spatial resolution is still limited by the spatial resolution of each satellite as explained above.
3.2.7 Coherence
As noted earlier it is not possible to produce interferograms if the signals from the two
acquisitions are not coherent. Coherence from an interferometric scene can be investigated
through coherence images that show the complex correlation between two SAR images using
both the amplitude and the phase information of the signals. Since various processes lead to
loss of coherence, the coherence image itself can be used for change detection. For alpine and
arctic areas these changes can be precipitation, variation in temperatures around freezing and
redistribution of snow. But also fast moving features like glaciers cause loss of coherence and
can therefore be detected (Weydahl 2001). However, the use of coherence images is limited to
detection of terrain deformation since no measurement of displacements is possible in the
coherence images. But they can serve as a tool for detecting deforming areas where noninterferometric techniques can be applied, either by remote sensing or field based
measurements.
3.2.8 Permanent scatterers technique
As noted above temporal decorrelation often prevents SAR interferometry from being
used for surface deformation. New techniques using stable natural reflectors, called
permanent scatterers (PSs) have therefore been developed. Pixels containing PSs will stay
coherent over long time intervals. The technique needs series of interferometric SAR images
but work well even for baselines longer than the critical ones. For pixels containing PSs, submeter DEM accuracy and millimetre terrain deformation can be detected. The reason is that
atmospheric phase contribution can be estimated and removed (Ferretti et al. 2000, 2001). The
technique relies on statistical methods and requires at least 25 -30 SAR images in order to
work. Subsidence, landslide deformation, crustal deformation, earthquake activity etc. have
been successfully mapped using the PS-technique (Ferretti et al. 2000, 2001, Dehls et al.
2002, Allievi et al. 2003, Chaabane et al. 2003, Colesanti et al. 2003, Lyons & Sandwell
2003).
35
3.3
Laser scanning
Using a laser sensor the distance between an aircraft and the ground or between a
terrestrial station and an object can be measured by the travel time of a reflected laser pulse.
This method is referred to as laser scanning or LIDAR (light detection and ranging). By a
number of such laser pulses a DTM can be generated (see Baltsavias (1999) and Wehr & Lohr
(1999) for an introduction to laser scanning). Just like photogrammetry laser scanners depend
on precise orientation parameters in order to achieve the demanded accuracy. This is usually
collected with INS (inertia navigation systems) and GPS (global positioning systems) aboard
the aircraft. The spatial resolution of DTMs acquired by airborne laser scanning is some
decimetres and the vertical accuracy in the centimetre-decimetre range.
Airborne laser
scanning has been used for mapping topography and surface elevation changes in glacial and
periglacial environments (Kennett & Eiken 1997, Favey et al. 1999, Baltsavias et al. 2001)
(Geist & Stötter 2003). For terrestrial based laser scanners a horizontal and vertical resolution
in the millimetre to centimetre range can be accomplished for short instrument-object
distances (Lim et al. 2005).
3.4
Comparison of the different techniques
The techniques presented in the previous sections operate at different resolutions and
have different accuracies, coverage, weather capabilities, costs and ranges. In the following
sections the different techniques are evaluated with regards to their applicability for
investigating the geomorphic processes described in chapter 2.
3.4.1 DTM differencing and surface elevation changes
The sensors and methods used for creating DTMs include terrestrial and aerial
photogrammetry, generation of DTMs from optical satellite data, air- and spaceborne InSAR
systems, and terrestrial and aerial laser scanning. Ground based surveying methods, which is
not elaborated in this thesis, includes tachymetry and GPS-profiling. Spaceborne LIDAR and
Radar altimetry systems can also be applied for DTM generation.
The main advantage of spaceborne systems is that they have a longer range than the
aerial and terrestrial based systems and are therefore possible to use for remote regions which
make up a large portion of the glacial and periglacial environments. Satellite data are usually
also cheaper than aerial or terrestrial campaigns. DTMs from satelliteborne sensors cover
36
larger areas in one scene, typically in swaths of some tens of kilometres to around 150
kilometres. But the vast coverage has an impact on the spatial resolution of the data sets,
which is in the range from some metres to 30 metres. For the mapping of smaller landforms
like rock glaciers this can be a great disadvantage since the whole landform is covered by
only a few DTM cells which make detailed studies impossible. This is an even greater
disadvantage for satelliteborne LIDAR and Radar altimeters since their large foot print only
make them suitable for DTM differencing of larger ice caps and ice sheets. The elevation
accuracy of DTMs produced from spaceborne systems is generally in the range of some tens
of metres and makes them suitable for the detection of large surface elevation changes only.
Airborne systems like aerial photogrammetry and laser scanning have a better
resolution and accuracy than the spaceborne but cover smaller areas. This fact increase the
potential for mapping smaller landforms and to measure surface changes in the decimetre to
metre range, but make regional studies more difficult. These systems are also more expensive
to utilize, unless there are possibilities of using old air photo archives. Depending on flying
height DTMs derived from pairs of aerial photographs cover areas from some hundred metres
to several kilometres.
Comparing InSAR data with optical data, the all weather and all day capability of
radar sensors make them very useful in glacial and periglacial regions and increase the
possibility for detailed monitoring of geomorphic processes. A disadvantage of InSAR is the
possibility of temporal decorrelation caused by changes in weather conditions or movements
causing loss of coherence. This makes InSAR difficult to use for topographic mapping of
crevassed glaciers and in some cases also for vegetated grounds. On the other hand DTMs
generated from optical remote sensing and photogrammetric techniques depend on a certain
contrast in order to work properly. Snow covered areas of glaciers are therefore often difficult
for photogrammetry. In this case laser scanning is a much better technique even though the
accuracy of photogrammetry in low contrast areas increase with novel sensors of higher
radiometric resolution (12-bit and 14-bit sensors compared to the older 8-bit systems often
used for scanning). For InSAR techniques the problem of microwave penetration of snow
covered surfaces also lead to low accuracies in the same areas where photogrammetry has to
deal with low contrast problems. In such areas laser scanning would be the preferred
technique since the laser signal is reflected at the snow surface. InSAR data are also affected
by geometrical effects like radar layover and radar shadowing for mountain topography due to
its side looking geometry. Back and forward looking optical sensors may also encounter
similar problems.
37
If the process investigated has surface elevation change in rates of millimetres to
centimetres and a very high spatial resolution is needed one has to apply terrestrial methods
like terrestrial laser scanning or terrestrial photogrammetry. Tachymetry or differential GPS
measurements could also be applied, but will lead to a lower number of points being
investigated. The satellite based permanent scatterers technique and also to some extent
InSAR can achieve the same vertical accuracies as the terrestrial methods. But the spatial
resolution of InSAR makes it suitable only for mapping horizontally large scale phenomena
like subsidence. The permanent scatterers technique depends on strong radar scatterers and
the number of suitable scatterers available usually leads to some distance between the
locations being measured. Terrestrial laser scanning and terrestrial photogrammetry are
therefore the only techniques suitable if a high spatial resolution is also needed. Some
processes like erosion of rock walls and cliffs resulting in retreat are not visible from satelliteand airborne platforms.
3.4.2 Measuring earth surface displacements from multi-temporal
imagery
As for generating DTM the different techniques available for tracking earth surface
movements also have their preferred area of application. As noted in the previous chapter
using InSAR can lead to problems of temporal decorrelation for crevassed and fast moving
glaciers, vegetated surfaces or for any surface due to a change in weather conditions. This will
also affect the ability to use InSAR for displacement mapping and for glacier velocity studies
limit the InSAR technique to the snow covered and homogenous areas of the glacier surface.
Optical feature tracking techniques on the other hand will not work in homogenous regions
since they depend on recognizable features and a certain level of contrast in the images. For
glacier velocity mapping this means that feature tracking is best suited for crevassed or
material covered areas of the glacier and act as a complementary technique to InSAR. The
accuracy of InSAR for displacement mapping is in the millimetre range which makes it far
more accurate than the one pixel accuracy of optical feature tracking techniques. The
temporal baseline for SAR scenes used for InSAR displacement mapping is limited by
temporal decorrelation, while optical feature tracking techniques must have a time interval
between the acquisitions that produce a displacement that is detectable with the one pixel
accuracy.
38
The 20 m ground resolution of satellite InSAR makes it unsuitable for mapping small
scale phenomena, like small rock glacier and also for mapping displacement with a high
degree of spatial variation. For such mapping feature tracking in aerial orthophotos are
superior. But it is worth noting that InSAR scenes are well suited for an area wide detection of
earth surface movement of small scale landforms and can serve as a first step for identifying
areas that can be investigated further by feature tracking with aerial orthophotos or field based
methods (Strozzi et al. 2004). Radar satellites have an all day and all weather capability while
feature tracking in optical imagery depends on a certain level of illumination restricting the
technique to imagery from clear summer days, making feature tracking unsuitable for
mapping inter-annual variation in displacements. Using feature tracking in orthorectified
imagery allow only the horizontal displacement components to be mapped. But techniques
have been developed for three dimensional displacement mapping using a rough DTM to
generate four “quasi-orthophotos” (one for each air photo used in the two models) and by
backprojecting the four matched points representing one identified feature in all orthophotos
into their respective original photographs (Kaufmann & Ladstädter 2003). Applying only one
INSAR scene will reveal a displacement vector towards the satellite. This vector can be
decomposed if some assumptions are made. For glacier one usually assumes surface parallel
flow. Using InSAR scenes from both ascending and descending orbits will produce velocity
vectors in three dimensions (Mohr et al. 1998).
This demonstrates that both InSAR and feature tracking in optical imagery are useful
techniques for tracking earth surface movements and that they are complementary. The
preferred technique will depend on surface characteristics, size of the landform, the level of
spatial and temporal detail needed, and the displacement rate dependant time scale.
The techniques of traditional and GPS surveying, SAR speckle tracking and InSAR
permanent scatterers, which are not used in this thesis, are to some extent all based on point
measurements. Traditional and GPS surveying can produce results of higher accuracy than
InSAR and feature tracking, but the number of points measured and possibility of high
temporal resolution is limited by its cost and labour intensiveness. But, surveyed
displacements are extremely valuable as ground truth for the remote sensing techniques
discussed earlier. SAR speckle tracking and InSAR permanent scatterers technique are limited
by the number of objects with high and stable radar backscatter properties. For glacier
surfaces they are scarce and for rock glaciers or slow moving land slides the density of
available measurement points would probably also lead to displacement mapping of low
39
resolution as was the case for InSAR. But for mapping large scale uniform displacements like
surface subsidence or uplift the technique of permanent scatterers are very well suited.
3.4.3 Remote sensing techniques in relation to geomorphic processes
Looking back at the concept of geomorphic work (Caine 1976) outlined in chapter 2.1
as vertical mass transfer we see that differencing of DTMs could be used for this purpose.
DTMs from air- or satelliteborne platforms can be used for measuring vertical changes and,
hence, also for calculating the volume and mass transfer of slope erosion and accumulation,
while terrestrial acquired DTMs can be applied for measuring erosion and accumulation in
steeper slopes and rock walls. By combining displacement fields from the earth surface
change tracking techniques described in this chapter with volume estimates from DTMs or
other methods, mass fluxes and geomorphic work of slow moving slopes can be assessed
more directly. These mass fluxes are also an expression of geomorphic work as introduced by
Caine (1976) based on the classic studies of Jackli (1957) and Rapp (1960b).
40
4 Summary of Research
This chapter gives an overview of the research carried out during the work with this
thesis. The different study areas are presented in the first section. An overview of the methods
used is given in the second section and the main results and analysis are presented in the last
section. In addition there is some further validation and applications of the results that are not
found in the papers.
4.1
Study areas
The research comprises four different projects carried out at the three different main
locations in Fig. 4; Iceland (1), Svalbard (2) and southern Norway (3). The study areas are
briefly presented below and more detailed descriptions can be found in each of the four
papers.
Paper I presents displacement measurements for three different locations in Iceland
(Fig. 5). Iceland is situated between 63q - 67q N and 14q - 24q W in the North-Atlantic.
Hóladalur and Fremri-Grjótárdalur make up the first location and are located close to Hólar
on the central part of the Tröllaskagi peninsula (65°40'N, 19°W) (1 in Fig. 5). The Tröllaskagi
peninsula is situated in northern Iceland between Skagafjörður in the west and Eyafjörður in
the east and made up of land ranging from 600 to 1,400 metres in altitude. Deep glacial
valleys and fjords cut into the mountainous peninsula. The studied landforms fill the valley
ends of Hóladalur and Fremri-Grjótárdalur and consist of a debris covered glacier and a
glacier-derived rock glacier in Hóladalur and a complex of glacier-derived rock glaciers in
Fremri-Grjótárdalur. All these landforms are situated between 900 and 1,200 m a.s.l. within
two cirques, surrounded by mountains and mountain plateaus reaching 1,300 m a.s.l.
Geophysical investigations and modelling indicate that the lower limit of mountain permafrost
is around 800 - 900 m.a.s.l. in the area.
The second location in Paper I is the moving debris accumulation at Almenningsnöf
between Kvígildi and Skriðnavik at the coast road to Siglufjörður on the northern tip of the
Tröllaskagi peninsula (66°10'N, 19°W) (2 in Fig. 5). The relief of the northern tip of the
peninsula is more alpine than at Hólar. The debris accumulation is situated below north and
west heading rock walls, between sea level and 260 m a.s.l. It comprises a coast of 2 km in
length and has an area of 1.9 km2.
41
Fig 4: Map of the North-Atlantic region, showing the research areas Iceland (1), Svalbard (2)
and southern Norway (3).
42
The third location is the debris-mantled slope above the Seyðisfjörður community (3
in Fig. 5). Here, the layer of moving debris does not show any specific landform connected to
surface movement in contrast to the two other investigated areas. The investigated area covers
around 1 km2 of the northeast-heading slope above the Seyðisfjörður community and ranges
from 30 to 340 metres in altitude. The mountains around the glacially deeply eroded
Seyðisfjörður reach more than 1,000 m a.s.l.
Fig. 5: Map of Iceland, showing the investigated areas of Paper I: Hóladalur and FremriGrjótárdalur (1), Almenningsnöf (2) and Seyðisfjörður (3).
Paper II and Paper IV are confined to Svalbard (Fig. 6). Svalbard is situated between
74q - 81q N and 10q - 35q E in the North-Atlantic, and has permafrost conditions down to sea
level (Liestøl 1977). Glaciers of different types cover 60% of the 66,000 km2 archipelago
(Hagen et al. 1993). Paper II presents retreat rates for two coastal cliff cliffs in the
Kongsfjorden area on Svalbard (4 in Fig. 6). The Kongsfjorden area is situated on the western
coast of Spitsbergen, the largest island of the Svalbard archipelago. The mean annual air
temperature (MAAT) in Ny-Ålesund located on the south side of Kongsfjorden is -6˚C and
43
Fig. 6: Map of Svalbard showing the location of the three glaciers investigated in Paper IV;
Isachsenfonna (1), Nordbreen (2), Akademikerbreen (3) and Kongsfjorden (4), the location of
the coastal retreat rate investigations in Paper II.
the mean annual precipitation 370 mm (Førland & Hanssen-Bauer 2000). The area has
continuous permafrost with measured depths ranging from 130 to 150 m near the shores of
Ny-Ålesund (Orvin 1944, Liestøl 1977) . The mean annual ground temperature (MAGT) is
believed to be about -5˚C based on measurements in a former coal mine (Orvin 1944). Paper
IV presents velocity field for three glaciers on Svalbard and all the glaciers used in this study
are also situated on Spitsbergen. The method was developed using a DInSAR scene of
Isachsenfonna (78q85'N, 13q11'E) (1 in Fig. 6). This glacier is situated close to Kongsfjorden
and the research settlement in Ny-Ålesund. The second glacier, Nordbreen (2 in Fig. 6), is
considerably smaller than the other two. It is situated in the north-eastern part of Spitsbergen
44
(79q38'N, 16q10'E). Nordbreen is an outlet from the Åsgårdsfonna ice cap. For location 3 in
Fig. 6, the name Akademikerbreen is used as a joint name for Transparentbreen, Opalbreen,
the lower part of Akademikerbreen and parts of Negribreen and Lomonosovfonna. Since
Akademikerbreen occupies the greater part of the InSAR scene, this name is used for
reference purposes. The centre of the scene is situated in the eastern part of Spitsbergen
(78q40'N, 18q40'E).
Fig. 7: Map showing the location of Jostedalsbreen in southern Norway and investigated area.
Paper III presents glacier velocity measurements and elevation surface change for
three glacier outlets on the eastern side of Jostedalsbreen in southern Norway (Fig. 7);
Nigardsbreen (61°41’N, 7°11’E), Baklibreen (61°40’N, 7°5’E) and Bergsetbreen (61°39’N,
7°4’E). Nigardsbreen is a valley glacier, 9.6 km in length, which drains a large area of the ice
cap and flow down into a deep U-shaped valley through three different tributary icefalls.
Baklibreen is a smaller and much steeper glacier. Bergsetbreen is a steep outlet draining the
area south of Baklibreen. It drops steeply into the valley end of Krundalen having a slope
angle of 27° which continuous over an area covering 1,000 vertical meters. The area of the
45
three investigated outlets is 48.2 km2, 3.19 km2 and 10.50 km2 for Nigardsbreen, Baklibreen
and Bergsetbreen, respectively. Nigardsbreen extends from 355 to 1,950 m a.s.l., Baklibreen
from 950 to 1,950 m a.s.l., and Bergsetbreen from 560 -19,60 m a.s.l. (Østrem et al. 1988).
4.2
Methods
The remote sensing methods used in this thesis can be divided into three; 1) automatic
cross-correlation matching of aerial orthophotos for displacement measurements, 2) automatic
generation and differencing of DTMs from digital photogrammetry for measuring glacier
surface elevation change and coastal cliff retreat rates, and 3) measuring glacier displacements
using height differentiated InSAR scenes. In this section the methods used will be described.
A more thorough description is found in the respective papers.
4.2.1 Coastal cliff retreat rates and glacier surface elevation change from
differencing DTMs
The general theory for automatic creation of DTMs is explained in section 3.1.2 and
the concept of differencing DTMs is explained in sections 3.1.4 and 3.4.1. The technique of
DTM differencing is used both in Paper II and Paper III. The ImageStation Automatic
Elevation module (ISAE) in Z/I Imaging is used for creating the DTMs. ISAE uses a
matching procedure similar to the one explained in section 3.1.2.
In Paper II DTMs acquired with terrestrial photogrammetry two years apart are used
to quantify retreat rates of two coastal cliffs in the Kongsfjorden area on Svalbard. As noted in
section 3.1.2 vertical accuracies of 1-3 parts per 10,000 of the camera/object distance rising to
10 to 30 parts per 10,000 for sparse ground control can be expected for DTMs from digital
photogrammetry. For the terrestrial photogrammetry in Kongsfjorden an accuracy of around
10-12 parts per 10,000 is assumed based on similar research by Pyle et al. (1997) and
Chandler et al. (2003), resulting in 15 mm accuracy for the 15 m camera/object distance the
first site and 7 mm accuracy for the 7 m camera/object distance at the second site. When
differencing two models with these accuracies, 21.2 mm and 9.9 mm accuracy can be
achieved for the two sites according to Eq. 3 in section 3.1.4. Since there is a two year interval
between the acquisitions the accuracy per year and lower limit of detectable retreat rates
would be 10.6 mm a-1 and 4.9 mm a-1. The relatively low accuracy compared to what is
possible with conventional aerial photogrammetry is explained by inaccuracies in estimating
46
ground control points being much larger relative to camera/object distance for close range
photogrammetry. The DTMs were differenced using the GRID module of ESRI’s ArcInfo
Workstation (ArcGIS version 8.3).
In Paper III DTMs from a variety of sources of three glacier outlets of Jostedalsbreen
are utilized for measuring glacier surface elevation change between the years 1984, 1993,
1997 and 2001. The two newest DTMs are generated with Z/I Imaging ImageStation. The
height accuracy of the 2001 and 1997 DTMs are assumed to be 0.8 m and 1.3 m (ca. 0.025%
of 3,100 and 5,300 m flying height), while both the 1984 and 1993 DTMs have an accuracy
closer to 5 m (Tab. 2). For a bare and well sunlit glacier surface of high optical contrast
having good ground control it is reasonable to expect an elevational accuracy of 2.5 parts per
10,000 of the flying height. The 1984 and 1993 DTMs have been interpolated from contour
lines of 10 m and 20 m equidistance respectively. Using again the Eq. 3 in section 3.1.4 it
gives accuracies between 1.5 and 7.1 m depending on which DTMs used for the
differentiation. As for the Paper II the GRID module of ESRI’s ArcInfo Workstation
(ArcGIS version 8.3) was used for the differencing.
Tab. 2: Data on the digital terrain models used. NVE is the Norwegian Water Resources and
Energy Directorate, Statens kartverk is the Norwegian Mapping Authority, and UoO is the
Department. of Geosciences, University of Oslo. Height accuracy for the 1997 and 2001
DTMs are calculated based on the flying height, an average terrain elevation of 850 m a.s.l.
with a height accuracy of ca 0.02 % of the flying height (above terrain).
Photo date
Resolution Flying
(m)
height
(m.a.s.l)
10. Aug 1984 25
6,300
08. Sep 1993 25
6,150
14. Aug 1997 5
6,150
29. Aug 2001 5
3,950
Height
accuracy
(RMS m)
5
5 (4-6)
1.3
0.8
Coverage
(glaciers)
Source
Nigardsbreen
All
Nigardsbreen
All
NVE
Statens kartverk
UoO
UoO
4.2.2 Glacier, rock glacier and slope displacements from crosscorrelation matching of orthophotos
The general concepts of cross-correlation matching of orthophotos are described in
section 3.1.5. In Papers I and III orthophotos are generated using a Z/I-Imaging
ImageStation digital photogrammetric workstation (DPW). The orthophotos are rectified
using DTMs produced in the same manner as explained in the previous section. Crosscorrelation matching of orthophotos done using the CIAS-software (Kääb & Vollmer 2000).
47
This software matches homologous points in two geo and co-referenced orthophotos of the
same area taken at different times using a similar method as those explained in section 3.1.5.
Points are selected as a regular grid or as single points in the orthophoto of time 1 and the
corresponding points are located in the orthophoto of time 2 using a double cross-correlation
function, which is more or less identical to the function presented in section 3.1.2. (Kääb uses
a slightly different notation and calls the function double cross-correlation (Kääb & Vollmer
2000, Kääb 2004)). The corresponding points are found by extracting a reference block, a
small window, from the orthophoto of time 1 around each point and the homolog of this small
window is searched for in a larger test area around the corresponding coordinate in the
orthophoto of time 2. A cross-correlation factor is calculated for each possible location of the
reference block within this test area. The location that yields the highest correlation factor is
taken to be the new position of the point in the orthophoto of time 2. If displacement has taken
place during the time interval between the two acquisitions, then the displacement is
measured as the horizontal distance between the two homologous points (i.e. this method does
not measure the vertical displacement). So instead of measuring x-parallax in aerial photos as
is the case for automatic DTM generation the method is used for displacement measurements
with orthophotos.
The accuracy of this method is found to be about the size of one pixel (Kääb &
Vollmer 2000). This accuracy comprises both the relative accuracy between the orthophotos
used including accuracy from the orientation of the photogrammetric models and accuracy of
the orthophoto generation depending on the accuracy of the DTM (see section 3.13), and the
cross-correlation matching technique itself. From the application of the CIAS method in this
thesis, it is worth noting that the direction of the measured displacement vectors deviated from
the main flow direction when entering areas of displacement corresponding to the size of one
pixel. Even though there is no direct accuracy assessment in this thesis this fact strengthens
the findings of one pixel as the lower limit for detecting displacements.
In Paper I air photos of different ground resolution and time intervals are used. The
accuracy of the photos used (i.e. ground resolution) is shown in Tab. 3. Depending on the
time interval between the air photo acquisitions this gives different accuracies for the annual
displacement, which can be taken as a limit for detectable annual displacements.
48
Tab. 3: Time interval between the air photo acquisitions, with total and per year accuracy for
the displacement measurements.
Location
Time
interval
Hóladalur
Almenningsnöf
Almenningsnöf
Seyðisfjörður
1985-1994
1977-1985
1985-1994
1964-1994
Accuracy
(m)
Accuracy
(cm a-1)
0.5
0.5
0.5
0.3
5.56
6.25
5.56
1.00
In Paper III the air photos have been acquired ten days apart giving a much shorter time
interval than for the study of Paper I. The ground resolution of the all orthophotos used in
Paper III is 0.5 metre, giving an accuracy for daily displacements of 5 cm/day.
The optimal size of the reference block and test area differs from study to study. The
size of the reference block must be large enough to contain enough information in order to
match recognisable features of the investigated surface and small enough to be distinct within
the test area. The test area has to be large enough to contain the reference block of photo 1
and the homolog of this block in photo 2 (i.e. twice as big as the displacement). The size of
both the reference block and the test area is determined by some initial testing. For Paper I
reference block sizes of 15x15 pixels were used for all locations. This size seems to be
sufficient for matching 0.3 and 0.5 m resolution photos of the coarse grained boulders found
on the landforms mapped in this study. Boulders ranging in diameter from one metre to a few
metres were used, for the manually selected points. The size of the test area was 50 x 50
pixels for the slow moving landforms and 100 x 100 pixels for the faster ones. Reference
block of 15x15 pixels were also used in Paper III, together with test areas of 100 x 100
pixels. This means that both the boulder surfaces of the landforms investigated on Iceland and
the crevassed bare glacier surfaces of the Jostedalsbreen outlets show distinct and matchable
features of the same size.
4.2.3 Measuring glacier displacements using height differentiated InSAR
scenes
In Paper IV the technique of height differentiated InSAR is used for measuring
glacier velocities for three glaciers on Svalbard. This means that the method developed use
motion only interferograms as input. In order to unwrap the DInSAR-scene, the discrete
fringe transitions had to be identified. First, a 3 x 3 median filter is applied twice to reduce the
49
characteristic speckle noise that is present in SAR and InSAR images. An edge-detecting
filter is applied after the median filtering. It is a derivative filter that uses the Prewitt-operator
to calculate the gradient of the pixel values in the directions of azimuth and range. There is
still some noise in the image after the Prewitt-filtering and the image is therefore threshold.
All pixels with values less than the threshold are set to 0, and the others are kept as they are.
A threshold of ca.150 has been shown to be suitable. The threshold image is then vectorized
with an existing algorithm in ArcInfo. Points that are closer than a certain distance are
connected with vectors. The resulting data have to be edited to some extent. This is similar to
the edge-segment linking performed by Lin et al. (1994). When the lines are edited, the
topology for the polygons that mark the transitions between the fringes can be created. The
polygons are numbered according to their relative position. This information is used in the
unwrapping of the InSAR image. Where half a wavelength (2.83 cm) is added to the medianfiltered InSAR-scene values inside fringe one, one wavelength is added inside fringe 2, etc.
The median-filtered InSAR-scene is used because it contains less noise and is believed to be
closer to the real phase differences than the original DInSAR scene. Wrongly classified pixels
are removed by statistical comparison with its neighbours. For reasons of simplicity, a median
filter is used for this task as well.
Since the scenes are already height differentiated, only one part of the expression in
Eq. 5 in section 3.2.4 is left, namely the contribution of glacier motion to the phase difference.
It could then be written as (Kwok & Fahnestock 1996):
'I mot
Where
&
v
4S
O
'U
4S &
&
v 'T ˜ r
O
is the surface ground velocity in the direction of the radar beam, ǻT the time
&
between the acquisitions (temporal baseline) and r the unit vector in direction of range. The
unwrapped values, explained in the proceeding part, represent the displacement towards the
satellite. A displacement calculated towards a satellite is of limited use from a glaciological
point of view. It is therefore common to assume surface parallel flow and decompose the
motion into the direction of surface slope.
Terrain slope is calculated together with aspect (i.e. direction of steepest surface
slope). This is done automatically with gradient filtering of a DTM. When the three
parameters of slope, aspect angle between glacier slope and radar range direction, and the
radar look angle have been calculated, together with the unwrapped interferogram, it is
50
possible to calculate the velocity, Vglac, in the flow direction of the glacier (K. Eldhuset,
personal communication 1998):
Vglac
Uw
cosD cosI sin T cosT sin D
( Eq. 6)
Uw is the displacement values in the unwrapped scene, D is the slope of the glacier, I is the
aspect angle between glacier slope and radar range direction, and T is the radar look angle.
Generally, there should also have been a 'T (temporal baseline) term in the denominator of
Eq. 6, but since it is tandem data and velocities are calculated in metres per day (m d-1), this
term is omitted. Since the glacier velocity in this study is defined as positive in the downward
direction of the steepest slope, the denominator of Eq. 6 has a plus sign differing from the
minus sign found in corresponding equations used in other studies, defining the velocity to be
positive upwards (Kwok & Fahnestock 1996). This difference is only due to the definition of
velocity direction and has no implications for the result. All the parameters are calculated for
every pixel at the glacier surface, resulting in a complete velocity field for the whole glacier.
4.3
Analysis and results
In this section the results from the four papers are presented. For some of the papers
additional analysis and verification not found in the papers are presented.
4.3.1 Coastal cliff retreat rates and glacier surface elevation change from
differencing DTMs
Coastal cliff retreat rates
The DTM differencing technique explained in section 4.2.1 is used in Paper II for the
coastal cliffs in the Kongsfjorden area and reveal an average retreat of 3.1 mm a-1 for site 1
and 2.7 mm a-1 for site 2 (see Paper II for figures). Both difference models show a spot like
erosion with maximum retreat rates of 12.5 cm a-1 and 4.2 cm a-1 for site 1 and site 2
respectively. It can be seen that several of the areas of greatest retreat at site 1 are found
around overhanging areas. The other retreat spots at both site 1 and site 2 are in areas where
the cliff is quite planar, and therefore show flaking rather than small rock falls. When
51
comparing orthophotos of site 1 from 2002 and 2004, they show removal of material in front
of the cliffs between the two acquisitions. The volume of the removed boulders along the 6.5
m distance with stereo coverage is estimated to be about 0.6 m3 based on photogrammetric
stereo measurements. Since the boulders are placed in front the cliff it is not possible to
measure them using DTM differencing.
The method of close up digital terrestrial photogrammetry seems suitable for creating
accurate DTMs of coastal cliffs and for calculating erosion rates in the order of millimetres to
centimetres. The retreat has a spot like appearance and the processes involved seem to be both
flaking and small rock falls. Compared to rock wall retreat rates found elsewhere on Svalbard
the rates are quite high. This could be due to other investigation being performed in noncoastal environments or the natural variation in the processes involved, making two years to
short a time interval of investigating coastal cliff retreat rates. Nevertheless, the retreat rates
found are within the range of world wide averages for limestone cliffs. Removal of weathered
material at the base of the coastal cliffs is considered to be important in the retreat process and
is documented through the multi-temporal photography obtained at one of the sites. About 0.6
m3 of material is removed from the 6.5 m wide investigation area at one of the investigated
sites.
Glacier surface elevation change
In Paper III surface elevation change for the three outlet glaciers Nigardsbreen,
Bergsetbreen and Baklibreen of Jostedalsbreen, southern Norway, have been calculated by
differencing the 5 m resolution Digital Terrain Model (DTM) obtained from the August 2001
air photos with two 25 m resolution DTMs from 1984 (Nigardsbreen) and 1993 (all three
outlets). The method is explained in section 4.2.1. The calculations show an average increase
in surface elevation of 22.1 m for Nigardsbreen between 1984 and 2001. Bergsetbreen and
Baklibreen show an increase of 3.2 and 14.3 m between 1993 and 2001. In addition a DTM
produced from air photos from 1997 was used for Nigardsbreen, making it possible to show
how the surface elevation has changed between the years 1984, 1993, 1997 and 2001. The
advance of the glacier snouts are also clearly visualized by the DTM differencing. The
accuracy of the measurements is shown in Tab. 4.
52
Tab. 4: Mean surface elevation change (m) from DTM differencing with the mean absolute
value in parentheses. For each calculation the glacier border of the last date has been used.
The accuracy (m) for each surface change calculation is also shown.
Nigardsbreen
Baklibreen
1984
- 2001
22.1
(22.2)
-
Bergsetbreen
Accuracy
5.1
1993
-2001
24.6
(24.6)
14.3
(14.3)
3.2
(6.6)
5.1
1984
-1993
-3.3
(5.7)
-
1993
–1997
19.3
(19.9)
-
1997
–2001
5.2
(5.6)
-
-
-
-
7.1
5.1
1.5
Assuming that Nigardsbreen and Bergsetbreen experienced the same changes in mass balance
during the late 1980s and early 1990s, it appears that they reacted in somewhat different
geometrical ways. From 1993 to 2001 Bergsetbreen only showed significant changes in its
lower parts while Nigardsbreen showed a general thickening in the whole ablation area. Both
showed thickening at the front and were advancing. The steep icefall of Bergsetbreen would
be able to transport a mass balance surplus faster down to the terminus than Nigardsbreen.
This explains why we only see a frontal change for Bergsetbreen whereas we see a change for
the whole investigated area at Nigardsbreen. From the surface elevation change it can
therefore be seen how the three different glaciers reacted to the highly positive winter balance
of the late 1980s and early 1990s. These differences can be explained by differences in slope,
glacier thickness, hypsometry and velocity.
4.3.2 Glacier, rock glacier and slope displacements from crosscorrelation matching of orthophotos
Glacier displacement
In addition to measuring surface elevation change, the study of Paper III aimed at
evaluating, for the first time, the applicability of high-resolution digital image-matching
techniques for deriving glacier surface velocities over short time intervals from repeat airphotos. The flow fields of two steep outlet glaciers, Bergsetbreen and Baklibreen, as well as
one valley glacier, Nigardsbreen, were therefore mapped using cross-correlation matching of
orthophotos based on air photos acquired 10 days apart in August 2001 (see section 4.2.2 on
the technique). Glacier displacements were calculated by matching homologous points in each
orthophoto using a regular grid with 10x10 m resolution. Displacements on Nigardsbreen
53
show daily rates of up to 1.34 m and agree well with GPS-measured velocities during the
same period (see Tabs. 5 & 6). The average velocities for the three glaciers ranged from 0.38
to 0.56 m d-1. See Paper III for velocity maps and interpolated velocity fields.
Tab. 5: Maximum and mean velocities measured by cross-correlation of the August 19 and 29
2001 orthophotos.
Nigardsbreen
Baklibreen
Bergsetbreen
Max (m d-1)
1.34
2.09
1.61
Mean (m d-1)
0.56
0.38
0.53
Tab. 6: Comparison of GPS measured stake displacements between June 22nd and August
22nd 2001 and between August 22nd and September 19th (Tønsberg 2003), and the average of
velocities obtained by cross-correlation of orthophotos acquired August 19th and 29th 2001 in
a 100 m radius circle around the stake positions. The position of the GPS measured stakes are
shown in Fig. 7 in Paper III.
GPS
Point
1
2
3
4
5
6
7
8
GPS
Velocity
Orthophoto
velocity 06–08.2001
-1
-1
(m d )
(m d )
0.31
0.35
0.61
0.60
0.57
0.59
0.60
0.61
0.62
0.60
0.66
0.67
0.70
0.70
0.77
0.80
Average deviation
Difference
Ortho and
GPS
velocity (%)
-12.9
1.6
-3.5
1.7
3.2
-1.5
0.0
-3.9
3.5
GPS
Velocity
08-09.2001
-1
(m d )
0.29
Difference
Ortho and
GPS
velocity (%)
6.5
0.53
0.54
0.55
0.60
0.65
0.73
7.0
10.0
11.3
9.1
7.1
5.2
8.0
The results are also compared to historic velocity measurements based on both terrestrial
photogrammetry and surveying (Tab. 7). Compared to the velocity reports from 1937 -1969,
as pointed out by Tønsberg (2003), there appears to have been an increase in velocities on the
tongue in later years. Exceptions to this are the velocities of point 5 and 6 from 1937 and
point 7 and 8 from 1949 and 1950 with measurements being close to the ones from 2001 (see
Fig. 7 in Paper III for point location).
54
Tab. 7: Comparison of glacier velocities (m d-1) conducted by terrestrial photogrammetry in
1937 (Pillewizer 1950), by Liestøl in 1949, 1951, 1953 and 1961 (Østrem et al. 1976),
trigonometric stake measurements in 1968/69 (Nielsen 1970), and the GPS and orthophoto
measured ones from 2001 presented here (Tønsberg 2003).
GPS
points
2&3
5&6
7&8
1937
1949 1951
1953
1961
1968/69
0.32
0.63
0.51
0.32
0.73
0.80
0.62
-
0.11
0.37
0.56
0.26
0.76
0.43
2001
GPS
0.60
0.64
0.75
2001
Ortho
0.59
0.64
0.74
Additional analysis
The result from the matching of multitemporal orthophotos resulted in a densely
measured velocity field for the three glaciers. It was therefore tried to use the technique
introduced by Nye (1959) for strain rate calculation from surveyed glacier velocities, and used
by Kääb & Funk (1999) on glacier velocity data from analytical photogrammetry. But a great
portion of the resulting longitudinal strain rates calculated were several order of magnitudes
greater than the upper limit of 10-2 a-1 for temperate glacier stated by Paterson (1994). The
reason for this is that small errors in the velocity based on photographs taken 10 days apart
would have a great impact on the per year values of longitudinal strain rate. The matching
routine depends on features like crevasses to succeed. This means that most of the measured
velocities are based on crevasses identified in the two orthophotos. The opening and closing
of crevasses during a 10 day period is probably not representative for the per year values of
longitudinal strain rate and leads to unrealistic values. Kääb & Funk (1999) use air photos
acquired at a one year interval. It was also tried to perform different sorts of filtering of the
original velocity data with no success.
Rock glacier displacement and slope displacement
In Paper I three areas of different active coarse debris slope processes are investigated
in Northern and Eastern Iceland. Surface displacement of some glacier derived rock glaciers
and a debris-covered glacier in the Hólar area, a moving debris accumulation close to
Siglufjörður and a debris layer in Seyðisfjörður are measured. The displacement fields are
obtained by cross-correlation matching of multi-temporal orthophotos as explained in section
4.2.2. Orthophotos are generated using various series of air photos from 1964 to 1994. The
results are analyzed and used for rough surface age estimates. In addition, type and cause of
movement are discussed. The velocity of the debris covered glacier and the rock glaciers in
the Hólar area average from 0.14 to 0.67 m a-1. The debris accumulation at Almenningsnöf
55
close to Siglufjörður shows an average displacement of 0.19 m a-1 with a maximum value of
0.84 m a-1 (see Tab. 8). The displacements at Almenningsnöf agree well with displacement
surveyed by GPS by the Icelandic Road Authorities.
The measured velocities in
Seyðisfjörður, although using air photos taken 30 years apart, turned out not to be significant,
but the homogenous direction of the displacement vectors suggests that the debris is currently
creeping. Based on the surface age results all landforms in Hóladalur are suggested to have
developed during the late-Holocene cooling period, with ages from around 1500 via 3000 to
around 5000 years for the different landforms. These surface age estimates coincide with data
from moraine dating nearby and Holocene climatic development in the North Atlantic region.
Tab. 8: Measured velocities at the three locations Hóladalur, Almenningsnöf and
Seyðisfjörður. The numbers after the names refer different landforms. See Paper I for
location.
Location
Hóladalur 1
Hóladalur 2
Hóladalur 3
Hóladalur 4
Hóladalur 5
Hóladalur 6
Hóladalur 7
Hóladalur 8
Hóladalur 9
Hóladalur 10
Hóladalur 11
Almenningsnöf
Almenningsnöf
Almenningsnöf 1
Almenningsnöf 1
Almenningsnöf 2
Almenningsnöf 2
Almenningsnöf 3
Almenningsnöf 3
Almenningsnöf 4
Almenningsnöf 4
Seyðisfjörður
Time period
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1985-1994
1977-1985
1985-1994
1977-1985
1985-1994
1977-1985
1985-1994
1977-1985
1985-1994
1977-1985
1985-1994
1964-1994
Average
velocity
(m a-1)
0.37
0.19
0.14
0.33
0.27
0.46
0.54
0.67
0.67
0.58
0.57
0.19
0.19
0.38
0.58
0.29
0.41
0.15
0.13
0.20
0.17
0.02
Maximum
velocity
(m a-1)
0.84
0.39
0.26
0.54
0.44
0.66
0.74
0.72
0.70
0.74
0.89
0.67
0.84
0.62
0.84
0.67
0.80
0.39
0.38
0.51
0.52
0.13
The displacement results from Hóladalur suggest that there is additional creep taking place in
the surface of the debris-covered glacier and the glacier derived rock glaciers. Based solely on
the displacements it is not possible to conclude on the cause of this additional creep. The
displacement results from the moving debris accumulation at Almenningsnöf close to
56
Siglufjörður agree well with existing GPS surveyed displacements and add more knowledge
on the dynamics of the landform. The homogenous movement for most of the debris
accumulation at Almenningsnöf could be due to movement along a sliding plane or
deformation within the debris itself; however, it is difficult to interpret the cause of movement
from the displacements and topography alone. An exception of that is the fast moving area 1
(Fig. 4 in Paper I). High and uniform velocities in this area (Fig.11 in Paper I) indicate that
this part of the debris accumulation is moving independently from the rest. The homogenous
displacement indicates some kind of slide movement along a horizon within the mass. Since
the surface is quite planar and does not show any vertical movement one can assume that the
slip surface also is planar. This indicates a translational slide (Cruden & Varnes 1996, Dikau
et al. 1996), and considering the high debris thickness (Guðmundsson 2000) it is assumed that
this area is a slow moving slab slide (Dikau et al. 1996). The assumption of a translational
slide is also strengthened by the fact that the movements are not constant confirmed by GPS
measurements of the Icelandic Road Authority.
Due to the uncertain origin of this landform it is difficult to relate the results of the
streamlines travel times to surface age. At Seyðisfjörður the displacement rates during the
investigated 30 year period were too small to give significant results taking in account the
accuracy of the method, but their uniform direction imply that movement takes place. The
results of this work show that cross-correlation of orthophotos is a valuable tool for
determining the surface displacement field of selected creeping slope landforms and that the
data can be used for estimating surface ages for some of these landforms.
Additional validation and analysis
After Paper I was submitted and accepted, points established and GPS-surveyed on
the landforms in Hóladalur and Almenningsnöf during the 2003 field campaign has been remeasured with GPS in 2005. The results of these measurements are shown in Figs. 8 and 9. In
Tabs. 9 and 10 the yearly displacements from the 2003 and 2005 GPS measurements are
compared to the yearly displacements in the vicinity found by cross-correlation matching of
the multitemporal orthophotos in Paper I.
Point 1985-1994 G P S
1
0.27
0.24
2
0.21
0.24
Tab. 9: Velocities from GPS measurements in August 2003
and August 2005 together with the average velocity from the
cross-correlation matching of orthophotos within a 100 metre
radius around the GPS measured points for landform nr. 1 in
Hóladalur (Fig. 2 in Paper I). All velocities are in m a-1.
57
Fig. 8: Location of the GPS points on the debris covered glacier in Hóladalur used for
comparison of the displacements found by cross-correlation of orthophotos. The circles show
the 100-metre radius around the GPS from which displacements found in the orthophotos are
compared.
Fig. 9: Location of the GPS points on the moving debris accumulation at Almenningsnöf used
for comparison of the displacements found by cross-correlation of orthophotos. The circles In
58
Hóladalur there is a fairly good agreement between the velocity measured by GPS from 2003
to 2005 and the displacement found in the orthophotos of 195 and 1994. This may indicate
that the velocity of the debris covered glacier is fairly constant. show the 75-metre radius
around the GPS from which displacements found in the orthophotos are compared.
Point
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1977-1985
0.30
0.21
0.21
0.27
0.25
0.36
0.21
0.27
0.40
0.33
0.18
0.28
0.21
0.17
0.31
0.24
0.15
0.40
0.39
0.17
0.14
0.39
0.34
0.13
0.30
0.22
0.21
1985-1994
0.29
0.29
0.16
0.30
0.24
0.31
0.14
0.24
0.42
0.35
0.13
0.32
0.23
0.09
0.43
0.20
0.13
0.55
0.54
0.19
0.10
0.18
0.54
0.07
0.47
0.23
0.17
GPS
0.10
0.05
0.06
0.12
0.07
0.12
0.04
0.03
0.05
0.05
0.04
0.06
0.04
0.02
0.06
0.07
0.03
0.75
0.76
0.05
0.04
0.60
0.57
0.04
0.30
0.19
0.17
Tab. 10: Velocities from GPS measurements in
August 2003 and August 2005 together with
the average velocity from the cross-correlation
matching of orthophotos within a 75 metre
radius around the GPS measured points for the
debris accumulation at Almenningsnöf (Fig. 4
in Paper I). All velocities are in m a-1.
For the velocities measured by GPS from 2003
to 2005 at Almenningsnöf the agreement is not
that good. For most of the slow moving areas
the GPS measured velocities are considerably
lower than the ones found by cross-correlation
matching of orthophotos from 1977 to 1985
and 1985 to 1994. On the other hand, the fast
moving parts (GPS points 17, 18, 21 and 22)
the velocities found from the orthophoto
matching are lower than the GPS measured
ones. Since the GPS measurement started 9
years after the last air photo was acquired it is
difficult to know if the difference is related to a
change in the dynamic properties of the slope
accumulation or inaccuracies of the procedure leading to the matching of orthophotos.
However, the GPS measurements show the same motion pattern as the found by the
orthophoto matching, with remarkable higher displacement rates in the area 1 in Fig. 4 in
Paper I.
4.3.3 Glacier displacements from height differentiated InSAR scenes
The main issues of Paper IV were to develop a method for unwrapping existing
DInSAR data of three Svalbard glaciers, to estimate glacier velocities from the unwrapped
DInSAR data, and to decompose these velocities into the direction of glacier surface flow
59
using a photogrametrically derived DEM. All these tasks were performed using the ESRI GIS
software Arc and no interferometry software was used. The glaciers investigated in this study
are Isachsenfonna, Nordbreen and Akademikerbreen (Fig. 6). Isachsenfonna is the only one of
these for which previous velocity data exists and the InSAR scenes of this glacier were
therefore chosen to develop the method.
The method described in section 4.2.3 has been used to calculate the surface velocities along
the surface-flow direction for the three glaciers at Svalbard in Paper IV. Some data from the
results are presented in Tab. 11. For maps of the velocity fields see Paper IV.
Tab. 11: Data from the velocity calculations showing maximum (Vmax) and average (Vaver)
velocities, the calculated average terrain slope and average aspect (i.e. angle between range
and the direction of steepest slope), and error of calculated glacier velocities caused by
relative errors in the DEM (Etopo).
Glacier
Isachsenfonna
Isachsenfonna
Akademikerbreen
Nordbreen
Nordbreen
Nordbreen
Location
79°N13°E
79°N13°E
79°N19°E
80°N16°E
80°N16°E
80°N16°E
Date Vmax Vaver Slope
(ERS1/ERS2) (m/d) (m/d) (deg)
10/11.10.96 0.42 0.23 0.46
05/06.04.96 0.42 0.18 0.45
20/21.10.95 0.41 0.07 1.19
27/28.09.95 0.36 0.17 1.02
05/06.04.96 0.36 0.13 1.14
10/11.05.96 0.33 0.10 1.07
Aspect Etopo (%)
(deg)
128.0
2.3
128.0
0.9
176.0
3.1
43.7
9.8
42.4
1.2
17.0
6.4
Generally, all the velocity maps in this paper show velocity patterns that conform to
glaciological theory; velocity increases with surface slope and decreases with increasing
width. Thus the glaciers show extending and compressive flow due to changes in surface
slope and width as expected. These factors overrun the climatically ones and therefore there is
no good correlation between the position of the equilibrium line and the maximum velocities.
The greatest change in velocity is found along the edges of the glacier, since the edges exert a
drag on the glacier. For Nordbreen, two late winter scenes and one autumn scene were
chosen. The highest velocities are found in the September 1995 scene, which also has a higher
mean velocity than the other scenes. This could indicate higher velocities during summer due
to more meltwater. Since the scene was acquired in late September in the Arctic, this is
probably not the case. Hence, there has to be an alternative explanation for the increase. The
autumn scene was also acquired the year before the spring scenes, so the explanation could
60
equally be one other than seasonal variation. To be able to determine a seasonal variation, a
scene from July or August is needed together with a better determination of the glacier border.
The results from Paper IV show that glacier velocities on Svalbard can be calculated
with DEMs and height-differentiated tandem InSAR data by the use of modules in ESRI’s
ArcInfo software. A semi-automatic algorithm is developed in order to calculate glacier
velocities in the flow direction, resulting in maps of the velocity fields for the glaciers. The
accuracy is greatest when the angle between direction of slope and range is small. The results
seem to be reliable and in agreement with other ground and satellite-borne observations.
61
5 General Discussion
5.1
Technique validation and applicability
Field validation or ground truth of remote-sensing techniques is of course important.
However, when applying multi-temporal methods using photographs acquired several decades
apart, validation of the velocities found for 10- 20 year periods is often is impossible to
achieve. But for Papers I there is surveying and GPS measurements of the displacements
covering the time period investigated with the matching of multi-temporal orthophotos for
parts of the investigated areas. For Paper III there are GPS measured velocities for the whole
summer covering the time interval between the two air photo campaigns as well, but no
measurement of the 10-day period only. For Paper IV there are GPS measurements of some
stakes used for mass balance measurements in the lower part of the area investigated. But
these are annual displacement measured and not directly comparable to the one day
displacement data obtained from InSAR tandem since due to the possibility of seasonal
variations. But nevertheless, all the surveying and GPS data show displacements quite similar
to the ones measured by the techniques used in this thesis. Both the technique of using crosscorrelation of orthophotos and height differentiated InSAR scenes for measuring displacement
have also been thoroughly validated in respect to their precision and accuracy by many other
authors (see chapter 3).
There are no direct field validations performed for the locations used for the DTM
differencing in this study either (Papers II and III). This would also have been more difficult
to orchestrate than the surveying required for displacement validation. For glacier surface
change measurement GPS profiling or laser scanning could have been used of the latter years.
For the coastal cliff retreat measurements on Svalbard terrestrial laser scanning could have
been used for validation as Lim et al. (2005) has shown for similar studies. Using sediment
traps as (Prick 2004) could have been used an indirect way for validating the coastal cliff
retreat rates. But this would have required frequent and year round inspection of the field
sites. For the glacier surface elevation change study mass balance and length change data
from the Norwegian Water Resource and Energy Directorate (NVE) (Kjøllmoen 2001, 2003)
serve as an indirect way of controlling the large scale findings, but not as a validation of the
method it self. But as is the case for the techniques used for displacement mapping there
63
exists several precision and accuracy studies of DTMs derived from digital photogrammetry
that are relevant to the study performed in this thesis. (See chapter 3).
Cross correlation matching of orthophotos for mapping slow mass movements can not
compete with GPS measurements when it comes to the accuracy alone. But, as shown in this
thesis, if a sufficiently long time interval is chosen between the air photo data used for the
displacement mapping in order to overcome the accuracy, the accuracy is no longer a
problem. This could also be turned into an advantage by using the huge post World War II air
photo archive available even for many periglacial areas. This makes it possible not only to
assess present displacement rates but also to get longer time series to look for possible
temporal changes in the displacement rates, caused by climatic changes or changes in other
factors controlling the movement. When a set of pass points has been acquired in field these
can, if wisely chosen be used for orientation of photogrammetric models based on future air
photos as well as historic photos. The availability of on flight GPS and inertia navigating
systems (INS) will also decrease the need for field based pass point campaigns in the future.
All together these characteristics make the cross-correlation matching of orthophotos a
feasible technique for monitoring displacements. Some potential areas of natural hazards
could be monitored in this way, making it possible to cover larger areas than what is
achievable in an efficient way with GPS monitoring.
Another great advantage of using cross-correlation matching of orthophotos is that the
number of resulting measurement points by far outnumbers what can be obtained by
traditional surveying or GPS. This is desirable when it comes to analyzing cause and type of
movement and as input and validation for models dealing with the mechanical flow properties
of both ice and soil. As shown in the next section, having a near complete displacement field
can be used for deriving several parameters important in both glaciology and process
geomorphology.
In this thesis it is also shown, as by many other authors, that digital photogrammetry
can be utilized for automatically creating DTMs at very different scale. Very precise DTMs of
high resolution can be generated from terrestrial photography, while coarser DTMs covering
larger areas may be created from air photos. This means that the digital photogrammetry to
some extent is scale independent, and that resolution is restricted by the limits of
photographing techniques. This scale dynamic property makes the technique suitable for
studies in geomorphology, since one is able to get the precision needed to investigate a certain
process by choosing a proper camera to object distance for the investigations. If one includes
DTMs created from optical satellite data in to photogrammetry, the camera object distance
64
would vary from around 800 km for satelliteborne platforms to a few meters for terrestrial
applications. As stated before, cross-correlation matching for displacement mapping was first
done using satellite scenes from optical sensors. This means that this technique is dynamic
when it comes to scale in the same manner as digital photogrammetry for creating DTMs
since it also depend on the scale of the input image data. This is not the case for radar
interferometry to the same extent due to the complexity involved with operating radar system
compared to photographical ones. Even though, interferometry has been used with luck for
satellite- and airborne platforms, a swell as ground based ones. The availability of high
resolution optical satellite images down to 0.6 m for Quickbird sensor and the advent of
digital aerial cameras capable of obtaining resolutions down 3 cm (Leberl et al. 2003) should
further widen the scales at which both DTMs can be generated and cross-correlation matching
can be performed, and further increase the applicability within geomorphic studies. Novel
digital aerial camera technology also has an improved radiometric resolution, usually 12 bit
compared to the traditional 8 bit resolution of older satellite systems and scanners. The
enhanced radiometric resolution will increase both the accuracy and possibility of crosscorrelation matching due to greater contrast in the imagery.
It is also worth noting again that for displacement mapping the two techniques used in
this thesis can be regarded more as complementary techniques rather than competing ones. As
seen in Paper IV glacier velocity mapping by InSAR only work for homogenous snow
covered areas and cannot be used in crevassed and rapidly deforming areas. On the other hand
this is exactly the areas where cross-correlation matching of orthophotos (or optical satellite
images) work well. Another issue is the coverage and resolution. Satelliteborne sensors are
more feasible for displacement mapping of vast areas and areas of low spatial variation of
displacements leading to a reduced need for the detail obtained by cross-correlation matching
of orthophotos.
For slow moving processes like the slopes investigated in Iceland in this thesis, a time
interval of several years is needed in order to get the required accuracy. This means that
processes showing seasonal variations, like what is proposed for the debris accumulation on
Almenningsnöf, the seasonal variation cannot be identified using cross-correlation of
orthophotos. This would require more or less continuous GPS surveillance or the use of
several InSAR scenes for interferometry or permanent scatterers measurements.
65
5.2
Geomorphic processes and analysis
5.2.1 Flow lines for travel time and age estimates
In Papers I and III the velocity measurements obtained by cross-correlation matching
of orthophotos have been used to derive surface flow lines for the investigated glaciers, rock
glaciers and slope accumulations. For a particle following this flow line through the velocity
field a travel time could then be estimated. For the moving landforms in Iceland in Paper I,
these travel times are used as rough age estimates for the landforms. This is relying on the
assumption of constant movement through the entire existence of the landform, and this
movement being similar to the velocities measured for the time period covered by the
multitemporal orthophotos. In addition the surface debris found at the front of the landforms
are assumed not to be overrun by the landforms (i.e. the material found on the front of the
landform is from the initiation of the landform). These assumptions are of course rather crude,
since the landform probably experiences lower velocities during its initiation and may have
endured higher velocities during epochs of its lifetime. This leads to something far from
accurate dating, and should be considered as rough age estimates only. Nevertheless, the age
estimates for the landforms in Hóladalur found by using this technique agree quite well with
moraine dating records in the vicinity and general Holocene climate development in the
North-Atlantic region. This rests on the assumption that the glaciers and rock glaciers of the
Hóladalur area were initiated during colder periods, when the glaciers producing the moraines
dated by Stötter et al. (1999) advanced. The post Little Ice Age genesis argued by Hamilton
and Whalley (1995) would imply velocities more than tenfold of what is found today, and
would also lead to very high weathering rates. It therefore seems to be more evidence
supporting an initiation during the second half of Holocene than during the Little Ice Age.
The use of flow lines and travel times for estimating age also rests on another
assumption, namely that all movement is surface parallel. For the debris cover of a glacier this
is valid since glaciological it is the travel times in the ablation area that is estimated. Rock
glaciers are also believed to show advection of material to a small extent. In contradiction to
glaciers which would show submerging movement in the accumulation area en emerging
movement in the ablation area, leading to velocity vectors not being surface parallel. This
means that surface velocities cannot be used directly for estimating travel time. All the areas
investigated in Paper III they are all below the equilibrium line and the travel times would
therefore be valid for a rock transported on the glacier surface, but not directly for a snow
66
particle. So, the travel times in Paper III should no be treated as exact travel times but they
still give an idea of the relative difference in how fast the glaciers would react to a mass
balance change. Debated glacier properties like the travel of kinematic waves are believed to
depend on the surface velocity. Since glacier experience melting in its ablation area the travel
times can off course not be used for age estimates.
5.2.2 Geomorphic work and retreat rates
Knowing the velocity field for landforms like the ones in Hóladalur in Paper I, one
could estimate the sediment flux by knowing the sediment thickness. In turn this could also be
converted into geomorphic work. Knowing the sediment volume and using the age estimates
one could also estimate the material production. In the case for the landforms in Hóladalur all
material is believed to originate from the retreat of the cirque headwall. The low velocity of
the debris-covered glacier would not lead to significant basal erosion and the apparently clear
melt water emerging from the landforms without pronounced sediments, indicate a relatively
stationary basal layer frozen to the subsurface. If the material currently making up the debris
cover and rock glaciers all originate from the headwall, one could therefore estimate the
average retreat rate of the headwall for the time period covered bye the age estimate.
The total area of the rock-glacier complex in Fremri-Grjótárdalur is 1.35 km2. If we
assume an average debris thickness of 3 m with a porosity of 40% we get a total debris
volume of 2.4 106 m3. The area of the headwall taken from the DTM is 1.3 km2, leading to a
total headwall retreat of 1.9 m. If we use the age estimate of 3,000 years obtained from the
streamlines in Paper I, the average retreat rate would be 0.6 mm a-1. This retreat rate estimate
is off course very crude since several approximations are made. Different estimates for both
debris thickness in the range 2 to 6 m and age estimates from 2.000 to 5.000 years would lead
to different retreat rates in the range of 0.25 to 1.9 mm a-1, nevertheless it illustrates how
detailed data on the velocity field, obtained through cross-correlation matching of
orthophotos, can be used to derive geomorphic process parameters. Assuming a generation of
the landforms after the Little Ice Age as Hamilton & Whalley (1995) with an age of 200 year
would lead to a retreat rate of 9.3 mm a-1 assuming a debris thickness of 3m.
The area of the debris covered glacier in Hóladalur (nr 1 in Fig. 4. Paper I) is 2.7 km2
and assuming a debris thickness of 2 m with a porosity of 40% would lead to a total volume
of 3.2 106 m3. Having a headwall area of 1.7 km2, this leads to a headwall retreat of 1.9 m as
well. Since the age of this landform is estimated to be 5,000 years the retreat rate would be 0.4
67
mm a-1. Again is should be stressed that these calculations should be treated more like
examples than accurate results. A debris thickness in the range of 1 to 5 m and age estimates
from 3.000 to 6.000 years imply retreat rates of 0.2 to 1.6 mm a-1. Assuming a generation of
the landforms after the Little Ice Age as Hamilton & Whalley (1995) with an age of 200 year
would lead to a retreat rate of 9.5 mm a-1 assuming a debris thickness of 3m.
Using Eq. 1 of section 2.1 and assuming the same volume as above (3.2 106 m3), a
density for the basalt of 2,900 kg m-3, and using the average velocity of 0.37 mm a-1 and
average slope of 8.4q, gives a total geomorphic work of 5.0 MJ a-1 and 1.8 MJ a-1 km-2.
Different values for sediment thickness in the range from 1 to 5 m would lead to values in the
range of 2.5 to 12.4 MJ a-1 and 0.9 to 4.6 MJ a-1 km-2 equating to 0.03 – 0.15 W km-2.
Calculating geomorphic work in this manner also rests on the assumption of no vertical
velocity differences within the debris cover, meaning that the debris cover only shows
velocity changes in the horizontal direction and that the surface velocity measured at each
point is valid for the whole vertical debris column beneath measured point. For a thin debris
cover assumed in this study this assumption would probably be valid, but when the debris
cover reaches a certain thickness there will also be deformation causing vertical velocity
differences.
Rock wall retreat based on volume and age estimates of both glacier-derived and talusderived rock glaciers are generally in the range of 1 to 5 mm a-1 (Barsch 1996, Humlum
2000). The ranges of retreat rates calculated above are therefore in the lower part of this
spectrum. This could also imply that the estimated debris volumes are too small or the age
estimates to short. However the erosion rates inferred from the 200 year age suggested
Hamilton & Whalley (1995) of almost 1 cm a-1 is very high by any standards, implying that
this age estimate is far to short. As stated in Paper I the age estimate of 200 years does no
comply with streamlines calculated from the present flow field of the rock glacier either.
Beylich (Beylich 2000) found present retreat rates in basaltic bedrock of 0.03 to 0.20 mm a-1
in the eastern fjord area of Iceland. This area is reaching down to sea level experience higher
temperatures than the periglacial environment of Hóladalur.
The estimated geomorphic work of 0.03 – 0.15 W km-2 for the debris covered glacier
found in this study is in the same range as values found by Rapp (1960b), Owens (1969) and
Caine (1974) for soil creep and solifluction of 0.166, 0.075 and 0.109 W km-2 for northern
Scandinavia, New Zealand and Colorado respectively.
The retreat rates found for the coastal cliff in the Kongsfjorden area on Svalbard in
Paper II are 3.1 mm a-1 for site 1 and 2.7 mm a-1 for site 2. Compared to the findings of Rapp
68
(1960a), André (1997), Berthling (2001) and Prick (2004) elsewhere on Svalbard the results
from this study show somewhat higher retreat rates. None of these localities are exposed to
marine processes and therefore not directly comparable with the results from this study. On
the other hand, Prick (2004) reports a rate of 1.9 mm a-1 for the sandstone and shale cliff wall
in Longyearbyen and this cliff is the most similar of earlier investigated localities because of
similarities in process and bedrock composition. The differences between the results found in
this study and the earlier works of Rapp (1960a), André (1997) and Berthling (2001) can be
attributed to differences in the effectiveness of the rock wall retreat processes involved
compared to coastal processes. The smaller retreat rate found by Prick (2004) may be
explained by differences in bedrock and that the location used in her study consists of bedrock
that is somewhat less susceptible to weathering.
It is reasonable to believe that the processes acting on walls and slope of different
types on Svalbard are all heavily influenced by the permafrost conditions and that frost
weathering is important for the rock fracturing. The effectiveness of cryogenic weathering is
depending on the amount of liquid water available and the temperature gradient in the ground,
as discussed by (Ødegård & Sollid 1993, Ødegård et al. 1995). The steep temperature gradient
is probably found in both headwalls and costal cliffs, but the availability of water through
snow melt is probably higher for coastal cliff than for steep headwalls. Together with marine
processes removing weathered material at the base of the cliff, this would lead to higher
retreat rates for coastal cliffs than for headwalls, given that climatic and geologic conditions
are the same.
By investigating more sites for retreat measurement, representative retreat values for
Svalbard’s entire coastline may be estimated and the total sediment-input to the sea estimated.
Since one should not extrapolated results from investigations at two sites for the entire
coastline, the sediment-input is not estimated in this study, however it is worth noting that the
retreat rates found on Svalbard represent a post-glacial retreat of 20-30 metre which
corresponds to the retreat achieved during some decades for soft sediment coasts along the
Laptev and Alaskan Beaufort sea (Are et al. 2005, Jorgenson & Brown 2005).
69
6 Conclusions
This thesis has evaluated different remote-sensing based methods for addressing
geomorphic processes in periglacial and glacial environments. The following technical and
scientific conclusion can be drawn:
x
All techniques used resulted in significant horizontal or vertical changes of the objects
investigated with measured values above the critical accuracy.
x
Cross-correlation matching of orthophotos generated by digital photogrammetry
results in high-resolution displacement fields for a variety of moving landforms, in the
range of centimetres to decimetres per year depending on observation time. For the
Icelandic landforms, measurements revealed active slope systems, with velocities for
the debris covered glacier and the rock glaciers in the Hólar area averaging from 0.14
to 0.67 m a-1. The debris accumulation at Almenningsnöf showed average
displacement of 0.19 m a-1 with a maximum value of 0.84 m a-1. The same technique
used for a 10-day interval at Nigardsbreen show daily movement of up to 1.34 m and
agree well with GPS-measured velocities during the same period. The average
velocities for the three investigated outlet glaciers of Jostedalsbreen ranged from 0.38
to 0.56 m d-1.
x
Velocity fields from cross-correlation matching of orthophotos can be used to estimate
surface travel times for the different landforms by generating flow lines. For the
landforms in northern Iceland, the travel times serve as age estimates, with estimated
values of 3.000 to 5.000 a. This agrees well with existing knowledge on the Holocene
climatic development in the region but is much older than what has been suggested by
earlier publications. This has important implications for understanding the Holocene
landscape development in the area.
x
By combining high-resolution velocity fields, travel times and rheological information
of the landforms, estimates for erosion rates, geomorphic work and headwall retreat
rates can be made. Within this thesis retreat rates of 0.2 to 1.9 mm a-1 were estimated
71
for the rock glacier setting in Hóladalur, Iceland. The estimated values are within a
realistic scale, however, data for validation are lacking.
x
Non-interferometric conventional GIS software can be used for glacier velocity
mapping with differentiated interferometric ERS-tandem scenes, by implementing a
semi-automatic algorithm using the ESRI Arc software. As this technique is valuable
for mapping glacier surface velocities outside the crevassed areas treated with crosscorrelation matching of multi-temporal aerial orthophotos, the two different
techniques are complementary.
x
Differencing of multi-temporal DTMs generated from aerial digital photogrammetry
reveal surface elevation changes for glacier outlets of Jostedalsbreen, southern
Norway that comply with length change data and mass balance change data observed
in the 1990s. DTM differencing show an average increase in surface elevation of 22.1
m for Nigardsbreen between 1984 and 2001, and 3.2 and 14.3 m between 1993 and
2001 for Bergsetbreen and Baklibreen.
x
The same technique of differencing high-resolution DEMs also reveal coastal cliff
retreat rates of 2.7 and 3.1 mm a-1 for two sites in Kongsfjorden, Svalbard using
terrestrial digital photogrammetry with camera/object distances of 7-15 metres
compared to the 3,100 to 5,400 metre flying heights used for the Jostedalsbreen glacier
outlets. This study implies terrestrial photogrammetry to be a valuable and costeffective tool for assessing coastal erosion. However, more field studies are needed,
covering longer time periods and different bedrock types, to give more sound
conclusions.
In general, the remote sensing-based techniques investigated in this thesis are valuable
in terms of accuracy and applicability for studying the earth surface in periglacial and glacial
environments through quantifying and analyzing processes that govern geomorphic
development. By using the different techniques, various observation time intervals and
observation distances, a variety of geomorphic processes may be investigated at the necessary
scale and accuracy.
72
7 Outlook and recommendations
This thesis has investigated the applicability of remote sensing techniques for
measuring and analysing earth surface processes in periglacial and glacial environments.
x
To gain further knowledge on these processes it is recommended to take advantage of
the huge post World War II air photo archive existing for many of the periglacial and
glacial environments in order to map displacements and surface elevation changes in
the order of centimetres to metres per year. Further, it can be investigated if the rates
of such processes have changed during the time period covered and to map areas that
have not yet been investigated.
x
In order to estimate geomorphic parameters like headwall retreat rates and geomorphic
work as illustrated in Chapter 5, field studies of sediment properties like thickness,
porosity and density should be done in a more detailed manner than what was possible
working with Paper I in addition to the displacement mapping.
x
For the regions investigated in this thesis a huge archive of ERS-1/ERS-2 tandem
InSAR scenes are also available from the mid 1990s. Utilizing this archive for
extensive mapping of glacier velocities in regions like Svalbard may serve as a unique
reference for future velocity measurements. For calving glaciers, knowledge on the
calving rate through interferometric velocity measurements could also be important
with regard to changing glacier mass balance in a changing climate.
x
The advent of new technology, like the improved radiometric and geometric resolution
of digital cameras, will lead to a higher accuracy of both aerial and terrestrial acquired
imagery. This will eventually lead to new applications of cross-correlation matching
of orthophotos and DTM differencing at scales that have been impossible until now.
For instance photogrammetry-based mapping of slow-moving processes like
solifluction would gain new insight of the processes involved. Improved radiometric
resolution will also ease automatic generation of DTMs by DPW in low contrast areas
like the accumulation area of glaciers due to higher contrast.
73
x
Both airborne and terrestrial laser scanning will lead to more DTMs being available,
and for processes like coastal cliff retreat to be mapped in a more detailed and
extensive way than what is possible with terrestrial photogrammetry. For a more
thorough and extensive knowledge of coastal retreat rates than what is achieved by the
two investigated sites presented in Paper II, terrestrial laser scanning of longer
portions of the coastline is recommended.
74
References
Ahnert, F. 1998. Introduction to Geomorphology. Arnold, London, p. 352.
Allievi, J., Ambrosi, C., Ceriani, M., Colesanti, C., Crosta, G.B., Ferretti, A. and Fossati.
2003. Monitoring slow mass movements with the permanent scatterers technique.
IGARSS'03, July 21-25, 2003 (Toulouse, France). 1, 215-217.
Al-Rousan, N. and Petrie, G. 1998. System calibration, geometric accuracy testing and
validation of DEM & orthoimage data extracted from SPOT stereopairs using
comercially available image processing systems. International Archives of
Photogrammetry & Remote Sensing 34:4, 8-15.
André, M.-F. 1997. Holocene Rockwall Retreat in Svalbard: A triple-rate evolution. Earth
Surface and Processes and Landforms 22:423-440.
Are, F.E., Grigoriev, M.N., Hubberten, H.-W. and Rachold, V. 2005. Using thermoterrace
dimensions to calculate the coastal erosion rate. Geo-Marine Letters 25:2-3, 121-126.
Armenakis, C. 1984. Deformation measurements from aerial photographs. International
Archives of Photogrammetry & Remote Sensing 25:A5, 39-48.
Baily, B., Collier, P., Farres, P., Inkpen, R. and Pearson, A. 2003. Comparative assessment of
analytical and digital photogrammetric methods in the construction of DEMs of
geomorphological forms. Earth Surface and Processes and Landforms 28:3, 307-320.
Baltsavias, E.P. 1996. Digital ortho-images - a powerful tool for the extraction of spatial- and
geo-information. ISPRS Journal of Photogrammetry & Remote Sensing 51:63-77.
Baltsavias, E.P. 1999. A comparison between photogrammetry and laser scanning. ISPRS
Journal of Photogrammetry & Remote Sensing 54:2-3, 83-94.
Baltsavias, E.P., Favey, E., Bauder, A., Boesch, H. and Peteraki, M. 2001. Digital surface
modelling by airborne laser scanning and digital photogrammetry for glacier
monitoring. Photogrammetric Record 17:98, 243-273.
Barsch, D. and Caine, N. 1984. The nature of mountain geomorphology. Mountain research
and development 4:4, 287-298.
Barsch, D. 1996. Rockglaciers. Springer, Berlin, p. 331.
Benn, D.I. and Evans, D.J.A. 1998. Glaciers and glaciation. Arnold, London, p. 734.
Bernstein, R. 1983. Image geometry and correction. In: Colwell, R.N. (ed.) Manual of Remote
Sensing. American Society of Photogrammetry, Falls Church, VA, 881-884 p.
Berthling, I. 2001. Slow periglacial mass wasting - processes and geomorphological impact.
Case studies form Finse, southern Norway and Prins Karls Forland, Svalbard. Series
of dissertations submitted to the Faculty of Mathematics and Natural Sciences.
University of Oslo. Oslo. Dissertation for the Degree of Dr. Scient
Beylich, A.A. 2000. Geomorphology, sediment budget, and relief development in Austdalur,
Astfirdir, East Iceland. Arctic, Antarctic, and Alpine Research 32:4, 466-477.
Bindschadler, R.A. and Scambos, T.A. 1991. Satellite-image-derived velocity field of an
Antarctic ice stream. Science 252:5003, 357-364.
Bishop, M., Bonk, R., Kamp Jr., U. and Shroder Jr., J.F. 2001. Terrain analysis and data
modeling for alpine glacier mapping. Polar Geography 25:3, 182-201.
Bloom, A.L. 1998. Geomorphology: A systematic analysis of late Cenozoic Landforms.
Prentice-Hall, Upper saddle river, New Yersey, p. 482.
Boulton, G.S. 1974. Processes and patterns of subglacial erosion. In: Coates, D.R. (ed.)
Glacial geomorphology. State University of New York, Binghampton, 41-87 p.
Boulton, G.S. 1975. Processes and patterns of subglacial sedimentation: a theoretical
approach. In: Wright, A.E. & Moseley, F. (eds.) Ice ages: Ancient and modern. Seel
House Press, Liverpool, 7-42 p.
75
Boulton, G.S. 1979. Processes of glacier erosion on different substrata. Journal of Glaciology
23:89, 15-38.
Boulton, G.S. and Jones, G.S. 1979. Stability of temperate ice caps and ice sheets resting on
beds of deformable sediment. Journal of Glaciology 24:90, 29-42.
Braithwaite, R.J. and Oleson, O.B. 1989. Calculation of glacier ablation from air temperature,
west Greenland. In: Oerlemans, J. (ed.) Glacier fluctuations and climatic change.
Kluwer, Dordrecht, 219-233 p.
Brecher, H.H. 1986. Surface velocity determination on large polar glaciers by aerial
photogrammetry. Annals of Glaciology 8:22-26.
Burtch, R. 1997. History of photogrammetry. Surveing Engineering Department, Ferris State
University:http://www.ferris.edu/htmls/academics/course.offerings/burtchr/sure340/n
otes/History.pdf. Accessed 23 August 2005.
Caine, N. 1974. The geomorphic processes of the alpine environment. In: Ives , J.D. & Berry,
R.G. (eds.) Arctic and alpine environments. Methuen, London, 721-748 p.
Caine, N. 1976. A uniform measure of subaerial erosion. Geological Society of America
Bulletin 87:137-140.
Campbell, J.B. 1996. Introduction to remote sensing. 2 ed. Taylor and Francis, London.
Chandler, J.H., Buffin-Bélanger, T., Rice, S., Reid, I. and Graham, D.J. 2003. The accuracy of
river a bed moulding/casting system and the effectiveness of a low-cost digital camera
for recording river bed fabric. Photogrammetric Record 18:103, 209-223.
Chaabane, F., Avallone, A., Tupin, F., Briole, P. and Maitre, H. 2003. Correction of local and
global tropospheric effects on differential SAR interferograms for the study of
earthquake phenomena. IGARSS 2003, July 21-25, 2003 (Toulouse, France). 5, 31993201.
Colesanti, C., Ferretti, A., Novali, F., Prati, C. and Rocca, F. 2003. SAR monitoring of
progressive and seasonal ground deformation using the permanent scatterers
technique. IEEE Transactions on Geoscience and Remote Sensing 41:7, 1685-1701.
Cruden, D.M. and Varnes, D.J. 1996. Lanslide type and processes. In: Turner, A.K. &
Schuster, R.L. (eds.) Landslides: investigation and mititgation. National Research
Council (U.S.) Transportation Research Board, 675 p.
Dehls, J.F., Basilico, M. and Colesanti, C. 2002. Ground deformation monitoring in the
Ranafjord area of Norway by means of the permanent scatterers technique IGARSS'02,
June 24-28, 2002 (Toronto, Canada). 1, 203-207.
Delacourt, C., Allemand, P., Casson, B. and Vadon, H. 2004. Velocity field of "La Clapière"
landslide measured by the correlation of aerial and QuickBird satellite images.
Geophysical Research Letters 31:15.
Dikau, R., Brunsden, D., Schrott, L. and Ibsen, M.-L. (eds.) 1996. Landslide recognition.
Identification, Movement and Causes, 1 ed. Chichester: John Wiley & Sons Ltd. 251
p.
Eckert, S., Kellenberger, T. and Itten, K. 2005. Accuracy assessment of automatically derived
digital elevation models from aster data in mountainous terrain. International Journal
of Remote Sensing 26:9, 1943-1957.
Eldhuset, K., Aanvik, F., Aksnes, K., Amlien, J., Andersen, P.H., Hauge, S., Isaksson, E.,
Wahl, T. and Weydahl, D.J. 1996. First results from InSAR processing on Svalbard.
ESA SP-406: Conference proceedings of Fringe'96 Zürich, Switzerland 30 September
- 2 October. European Space Agency, Munich, p. 376.
Etzelmüller, B. and Sollid, J.L. 1996. Long-term mass balance of selected polythermal
glaciers on Spitsbergen, Svalbard. Norsk Geografisk Tidskrift - Norwegian Journal of
Geography 50:55-66.
76
Etzelmüller, B. 2000. On the quantification of surface changes using Grid-based digital
eleveation models (DEMs). Transactions in GIS 4:2, 129-143.
Fahnestock, M., Bindschadler Robert, A., Kwok, R. and Jezek, K. 1993. Greenland ice sheet
surface properties and ice dynamics from ERS-1 SAR imagery. Science 262:5139,
1530-1534.
Favey, E., Geiger, A., Gudmundsson, G.H. and Wehr, A. 1999. Evaluating the potential of an
airborne laser scanning system for measuring volume changes of glaciers. Geografisk
Annaler 81A:4, 555-561.
Ferretti, A., Prati, C. and Rocca, F. 2000. Non-linear subsidence rate estimation using
permanent scatterers in differential SAR interferometry. IEEE Transactions on
Geoscience and Remote Sensing 38:2202-2212.
Ferretti, A., Prati, C. and Rocca, F. 2001. Permanent scatters in SAR interferometry. IEEE
Transactions on Geoscience and Remote Sensing 39:1, 8-20.
Finsterwalder, R. 1931. Geschwindigkeitsmessungen an Gletschern mittels Photogrammetrie.
Zeitschrift für Gletscherkunde für Eiszeitforschung und Geschichte des Klimas
19:251-262.
Forster, R.R., Jezek, K.C., Koenig, L. and Deeb, E. 2003. Measurement of glacier geophysical
properities from InSAR wrapped phase. IEEE Transactions on Geoscience and
Remote Sensing 41:11, 2595-2604.
French, H.M. 1996. The periglacial environment. 2nd ed. Longmann, London, p. 341.
Fryer, J.G., J.H., C. and Cooper, M.A.R. 1994. On the accuracy of heighting from aerial
photographs and maps; implications to process modellers. Earth Surface and
Processes and Landforms 19:6, 577-583.
Førland, E.J. and Hanssen-Bauer, I. 2000. Increased precipitation in the Norwegian Arctic;
true or false?. Climatic change 46:4, 485-509.
Gabriel, A.K. and Goldstein, R.M. 1988. Crossed orbit interferometry: theory and
experimental results from SIR-B. International Journal of Remote Sensing 9:857-872.
Geist, T. and Stötter, H. 2003. First results of airborne laser scanning technology as a tool fo
the quantification of glacier mass balance. EARSel Proceedings 2:1, 8-14.
Gens, R. and van Genderen, J.L. 1996. Review article: SAR interferometry -issues,
techniques, applications. International Journal of Remote Sensing 17:10, 1803-1835.
Goldstein, R.M. and Zebker, H.A. 1987. Interferometric radar measurements of ocean
currents. Nature 328:20, 707-709.
Goldstein, R.M., Zebker, H.A. and Werner, C.L. 1988. Satellite radar interferometry: Twodimensional phase unwrapping. Radio Science 23:713-720.
Goldstein, R.M., Engelhardt, H., Kamb, B. and Frolich, R.M. 1993. Satellite radar
interferometry for monitoring ice sheet motion; application to an Antarctic ice stream.
Science 262:5139, 1525-1530.
Gonzalez, R.C. and Woods, R.E. 1993. Digital Image Processing. Addison-Wesley Pub.
Company, Reading, MA.
Gorbunov, A.P. and Titkov, S.N. 1992. Dynamics if rick glaciers of the northern Tien Shan
and the Djungar Ala Tau, Kazakhstan. Permafrost and Periglacial Processes 2:29-39.
Graham, L.C. 1974. Synthetic interferometer radar for topographic mapping. Proceedings of
the IEEE 62:763-768.
Guðmundsson, Á. 2000. Frerafjöll og Urðabingir á Tröllaskaga (in Icelandic). University of
Iceland. M.Sc.
Haeberli, W., King, L. and Flotron, A. 1979. Surface movement and lichen cover studies at
the active rock glacier near the Grubengletcher, Wallis, Swiss Alps. Arctic and Alpine
Research 11:4, 421-441.
77
Haeberli, W. 1985. Creep of mountain permafrost; internal structure and flow of alpine rock
glaciers. Mitteilungen der Versuchsanstalt fuer Wasserbau, Hydrologie und
Glaziologie. 77. Eidgenoessische Technische Hochschule in Zuerich, Zurich,
Switzerland, p. 142.
Haeberli, W. and W., S. 1988. Aerophotogrammetrical monitoring of rock glaciers.
Proceedings of the 5th International Conference on Permafrost 1:764-769.
Haeberli, W., Hoelzle, M., Kaeaeb, A., Keller, F., Vonder, M.D. and Wagner, S. 1998. Ten
years after drilling through the permafrost of the active rock glacier Murtel, Eastern
Swiss Alps; answered questions and new perspectives. In: Lewkowicz, A.G. & Allard,
M. (eds.) Proceedings, 7th International conference on permafrost, Yellowknife,
Canada. Centre d'Etudes Nordiques, Université Laval, Quebec, 403-410 p.
Haeberli, W. 2000. Modern Research Perspectives Relating to Permafrost Creep and Rock
Glaciers. Permafrost and Periglacial Processes 11:290-293.
Hagen, J.O., Liestøl, O., Roland, E. and Jørgensen, T. 1993. Glacier Atlas of Svalbard and
Jan Mayen. Norsk Polarinstitutt Meddelelser. 129. Norsk Polarinstitutt, Oslo, p. 141.
Hall, K., Thorn, C.E., Matsuoka, N. and Prick, A. 2002. Weathering in cold regions: some
thoughts and perspectives. Progress in Physical Geography 26:4, 577-603.
Hallet, B. 1979. A theoretical model of glacial abrasion. Journal of Glaciology 23:89, 39-50.
Hallet, B. 1981. Glacial abrasion and sliding: their dependence on the debris concentration in
basal ice. Annals of Glaciology 2:23-28.
Hallet, B., Walder, J.S. and Stubbs, C.W. 1991. Weathering by segregation ice growth in
microcracks at sustained subzero temperatures: verification from an experimental
study using acoustic emissions. Permafrost and Periglacial Processes 2:283-300.
Hamilton, S.J. and Whalley, W.B. 1995. Preliminary results from the lichenometric study of
the Nautardálur rock glacier, Tröllaskagi, northern Iceland. Geomorphology 12:123132.
Hoen, E.W. and Zebker, H.A. 2000. Penetration depths inferred from interferometric volume
decorrelation observed over the Greenland ice sheet. IEEE Transactions on
Geoscience and Remote Sensing 38:6, 2571-2583.
Humlum, O. 1982. Rock glacier types on Disco Island, Central West Greenland. Geogr.
Tidsskr. 82:59-66.
Humlum, O. 2000. The geomorphic significance of rock glaciers: estimates of rock glacier
debris veolumes and headwakk recession rates in Western Greenland. Geomorphology
35:41-67.
Høgbom, B. 1914. Ûber die geologische bedeutung des frostes. In: Sjøberg, H. (ed.) Bulletin
of the Geological Institution of the University of Upsala. 294-306 p.
IAHS(ICSI), UNEP and UNESCO. 1998. Fluctuations of glaciers 1990-1995. In: Haeberli,
W., Hoelzle, M., Suter, S. & Frauenfelder, R. (eds.). International Association of
Hydrologic Sciences, Paris, 296 p.
IAHS(ICSI), UNEP, UNESCO and WMO. 2001. Glacier Mass Balance Bulletin No. 6. In:
Haeberli, W., Frauenfelder, R. & Hoelzle, M. (eds.). World Glacier Monitoring
Service, Zürich, 93 p.
Iken, A. and Bindschadler, R.A. 1986. Combined measurements pf subglacial water pressure
and surface velocity of Findelengletscher, Switzeland: Conclusions about drainage
system av sliding mechanism. Journal of Glaciology 32:110, 101-117.
Jackli, H. 1957. Gegenwartsgeologie des bündnerischen Rheingebietes; ein Beitrag zur
exogenen Dynamik alpiner Gebirgslandschaften. Beitraege zur Geologie der Schweiz,
Geotechnische Serie. 36. p. 135.
78
Jorgenson, M.T. and Brown, J. 2005. Classification of the Alaskan Beaufort Sea Coast and
estimation of carbon and sediment inputs from coastal erosion. Geo-Marine Letters
25:2-3, 69-80.
Joughin, I.R., Winebrenner, D.P. and Fahnstock, M.A. 1995. Observations of ice-sheet
motion in Greenland using satellite radar interferometry. Geophysical Research
Letters 22:5, 571-574.
Kaufmann, V. and Ladstädter, R. 2003. Quantitative analysis of rock glacier creep by means
of digital photogrammetry using multi-temporal aerial photographs: Two case studies
in the Austrian Alps Permafrost: Proceedings of the Eighth International Conference
on Permafrost (Zurich, Switzerland). 1, 525-530. A.A. Balkema Publishers.
Kennett, M. and Eiken, T. 1997. Airborne measurements of glacier surface elevation by
scanning laser altimeter. Annals of Glaciology 24:293-296.
Kenyi, L.W. and Kaufmann, V. 2003. Estimation of rock glacier surface deformation using
SAR interferometry data. IEEE Transactions on Geoscience and Remote Sensing 41:6,
1512-1515.
Kjøllmoen, B. (ed.) 2001. Glaciological investigations in Norway in 2000. Report no 2:
Norwegian Water Resources and Energy Directorate (NVE), Oslo, Norway. 122 p.
Kjøllmoen, B. (ed.) 2003. Glaciological investigations in Norway in 2002. Report no 3:
Norwegian Water Resources and Energy Directorate (NVE), Oslo, Norway. 92 p.
Kjøllmoen, B. (ed.) 2004. Glaciological investigations in Norway in 2003, Report no 4:
Norwegian Water Resources and Energy Directorate (NVE), Oslo, Norway. 98 p.
Kwok, R. and Fahnestock, M.A. 1996. Ice Sheet Motion and Topography from Radar
Interferometry. IEEE Transactions on Geoscience and Remote Sensing 34:1, 189-200.
Kääb, A., Haeberli, W. and Gudmundsson, G.H. 1997. Analysing the creep of mountain
permafrost using high precision aerial photogrammetry; 25 years of monitoring
Gruben rock glacier, Swiss Alps. Permafrost and Periglacial Processes 8:4, 409-426.
Kääb, A. and Funk, M. 1999. Modelling mass balance using photogrammetric and
geophysical data: a pilot study at Griesgletcher, Swiss Alps. Journal of Glaciology
45:151, 575-583.
Kääb, A. and Vollmer, M. 2000. Surface Geometry, Thickness Changes and Flow Fields on
Creeping Mountain Permafrost: Automatic Extraction by Digital Image Analysis.
Permafrost and Periglacial Processes 11:315-326.
Kääb, A. 2002. Monitoring high-mountain terrain deformation from repeated air- and
spaceborne optical data: examples using digital aerial imagery and ASTER data.
ISPRS Journal of Photogrammetry and Remote Sensing 57:1-2, 39-52.
Kääb, A., Huggel, C., Paul, F., Wessels, R., Raup, B., Kieffer, H. and Kargel, J. 2002. Glacier
Monitoring from ASTER Imagery: Accuracy and Applications EARSeL Proceedings.
LIS-SIG Workshop. Berne, March 11-13. in press.
Kääb, A. 2004. Mountain glaciers and permafrost creep. Research perspectives from earth
observation and geoinformatic technologies. Habilitation thesis. Department of
Geography, University of Zurich, p. 204.
Kääb, A. and Weber, M. 2004. Development of transverse ridges on rock glaciers: Field
measurements and labratory experiments. Permafrost and Periglacial Processes
15:379-391.
Kääb, A. 2005. Combination of SRTM3 and repeat ASTER data for deriving alpine glacier
flow velocities in the Bhutan Himalaya. Remote Sensing of Environment 94:4, 463474.
Kääb, A. and Reichmuth, T. 2005. Advance mechanisms of rock glaciers. Permafrost and
Periglacial Processes 16:187-193.
79
Kääb, A., Lefauconnier, B. and Melvold, K. in press. Flow field of Kronebreen, Svalbard,
using repeated Landsat7 and ASTER data. Annals of Glaciology 42.
Leberl, F., Gruber, M., Ponticelli, M., Bernoegger, S. and Perko, R. 2003. The Ultracam large
format aerial ditigal camera system Proceedings of The American Society For
Photogrammetry & Remote Sensing
Anchorage Annual Conference, Alaska, 5-9 May, 2003.
Lefauconnier, B., Hagen, J.O. and Rudant, J.P. 1994. Flow speed and calving rate of
Kongsbreen glacier, Svalbard, using SPOT images. Polar Research 13:1, 59-65.
Liestøl, O. 1977. Pingos, springs, and permafrost in Spitsbergen. Norsk Polarinstitutt Årbok
1975 7-29.
Lim, M., Petley, D.N., Rosser, N.J., Allison, R.J., Long, A.J. and Pybus, D. 2005. Combined
digital photogrammetry and time-of-flight laser monitoring for cliff evolution.
Photogrammetric Record 20:110, 109-129.
Lin, Q., Vesecky, J.F. and Zebker, A. 1994. Phase Unwrapping Through Fringe Line
Detection in Synthetic Aperture Radar Interferometry. Applied Optics 33:2, 201-208.
Lucchitta, B.K. and Ferguson, H.M. 1986. Antarctica: Measuring glacier velocity from
satellite images. Science 234:4780, 1105-1108.
Lyons, S. and Sandwell, D. 2003. Fault creep along the southern San Andreas from
interferometric synthetic aperture radar, permanent scatterers, and stacking. Journal of
Geophysical Research 108:B1, doi:10.1029/ 2002JB001831.
Massonnet, D., Rossi, M., Carmona, C., Adragana, F., Peltzer, G., Feigl, K. and Rabaute, T.
1993. The displacement field of the Landers earthquake mapped by radar
interferometry. NATURE 364:138-142.
Massonnet, D. and Feigl, K.L. 1998. Radar interferometry and its application to changes in
the earth's surface. Reviews of Geophysics 36:4, 441-500.
Matsuoka, N. 1990. The rate of rock breakdown by frost action: field measurements and a
predictive model. Earth Surface and Processes and Landforms 15:73-90.
Matsuoka, N. 1991. A model of the rate of frost shattering: application to field data from
Japan, Svalbard and Antartica. Permafrost and Periglacial Processes 2:271-281.
Melvold, K. 1992. Studie av brebevegelse på Kongsvegen og Kronebreen, Svalbard (in
Norwegian). Rapportserie i naturgeografi. Universitetet i Oslo, p. 60.
Michel, R. and Rignot, E. 1999. Flow of Glaciar Moreno, Argentina, from repeat-pass Shuttle
Imaging Radar images: comparison of the phase correlation method with radar
interferometry. Journal of Glaciology 45:149, 93-100.
Mikhail, E.M., Bethel, J.S. and McGlone, J.C. 2001. Introduction to modern photogrammetry.
Wiley, New York, p. 479.
Mohr, J.J., Reeh, N. and Madsen, S.N. 1998. Three-dimensional glacial flow and surface
elevation measured with radar interferometry. Nature 391:6664, 273-276.
Maalen Johansen, I. and Andersen, Ø. 1998. En innføring i digital fotogrammetri til
kartleggingsformål (in Norwegian). Norges Landbrukshøgskole, Institutt for kartfag,
Ås, p. 76.
Nansen, F. 1922. The strandflat and isostasy. Videnskapsselskapets skrifter I. Mat-Naturv.
klasse 1921. 2. p. 313.
Nielsen, C. 1970. Triangulering av bevegelsesretninger i 1969. In: Pytte, R. (ed.)
Glasiologiske undersøkelser i Norge 1969. Hydrologisk avdeling, Norges Vassdrags
og elektrisitetsvesen, 96 p.
Nielsen, N. 1979. Ice-foot processes. Observations of erosion on a rocky coast, Disko, West
Greenland. Zeithschrift für Geomorphologie 23:3, 321-331.
Nye, J.F. 1959. A method of determening the strain-rate tensor at the surface of a glacier.
Journal of Glaciology 3:25, 409-419.
80
Orheim, O. and Lucchitta, B.K. 1987. Snow and ice studies by thematic mapper and
multispectral scanner landsat images. Annals of Glaciology 9:109-118.
Orvin, A.K. 1944. Litt om kilder på Svalbard (in Norwegian). Norsk Geografisk Tidskrift Norwegian Journal of Geography 10.
Owens, I.F. 1969. Causes and rates of soil creep in the Chilton Valley, Cass, New Zealand.
Arctic and Alpine Research 1:213 - 220.
Paterson, W.S.B. 1994. The Physics of glaciers. Pergamon/Elsevier Science, Oxford, p. 480.
Pillewizer, W. 1950. Bewegungsstudien an Gletschern des Jostedalsbre in Süd-Norwegen.
Erdkunde 4:201-206.
Prick, A. 2004. Observations of rock temperatures and rock moisture variability in
Longyearbyen: Implications for cryogenic weathering and rock wall retreat rate
Abstracts of Pace 21: Permafrost and Climate in the 21st Century. Field Workshop
Longyearbyen, Svalbard, 8-13 September 2004 16.
Pyle, C.J., Richards, K.S. and Chandler, J.H. 1997. Digital photogrammetric monitoring of
river bank erosion. Photogrammetric Record 15:89, 753-764.
Rachold, V., Are, F.E., Atkinson, D.E., Cherkashov, G. and Solomon, S.M. 2005. Arctic
Coastal Dynamics – an introduction. Geo-Marine Letters 25:2-3, 63-68.
Rapp, A. 1960a. Talus slopes and mountain walls at Tempelfjorden, Spitsbergen - a
geomorphological study of the denudation of slopes in an arctic locality. Meddelanden
Uppsala Universitets Geografiska Institution. 55A. p. 96.
Rapp, A. 1960b. Recent developments of mountain slopes in Kärkeveagge and surroundings,
northern Scandinavia. Geografisk Annaler 42:71-200.
Rignot, E., Jezek, K.C. and Sohn, H.G. 1995. Ice Flow Dynamics of the Greenland Ice Sheet
from SAR Interferometry. Geophysical Research Letters 22:5, 575-578.
Rignot, E., Hallet, B. and Fountain, A. 2002. Rock glacier surface motion in Beacon Valley,
Antarctica, from synthetic-aperture radar interferometry. Geophysical Research
Letters 10.1029/2001GL013494, 29:June 2002.
Rignot, E., Rivera, A. and Casassa, G. 2003. Contribution of the Patagonia Icefields of South
America to sea level rise. Science 302:434-437.
Roer, I., Kääb, A. and Dikau, R. 2005. Rockglacier acceleration in the Turtamann valley
(Swiss Alps): Probable controls. Norsk Geografisk Tidskrift - Norwegian Journal of
Geography 59:2, 157-163.
Rogers, A.E.E. and Ingalls, R.P. 1969. Venus: Mapping of the surface reflectivity by radar
interferometry. Science 165:797-799.
Rolstad, C. 1995. Satellitt- og flybilder til bestemmelse av bredynamikk (in Norwegian).
Departement of Physical Geography. University of Oslo, Norway. M.Sc.-thesis
Rolstad, C., Amlien, J., Hagen, J.O. and Lundén, B. 1997. Visible and near infrared digital
images for determination of ice velocities and surface elevation during a surge on
Osbornebreen, a tide water glacier in Svalbard. Annals of Glaciology 24:255-261.
Rott, H. and Siegel, A. 1996. Glaciological Studies in the Alps and in Antarctica Using ERS
Interferometric SAR. ESA SP-406: Conference proceedings of Fringe'96 Zurich,
Switzerland, 30 September - 2 October. European Space Agency, Munich, p. 379.
Rott, H., Scheuchl, B. and Siegel, A. 1999. Monitoring very slow slope movements by means
of SAR interferometry: A case study from a mass waste above a reservoir in the Ötztal
Alps, Austria. Geophysical Research Letters 26:11, 1629-1632.
Scambos, T.A., Dutkiewics, M.J., Wilson, J.C. and Bindschadler, R.A. 1992. Application of
image cross-correlation to the measurement of glacier velocity using satellite image
data. Remote Sensing of Environment 42:177-186.
Schweizer, J. and Iken, A. 1992. The role of bed separation and friction in sliding over an
undeformable bed. Journal of Glaciology 38:128, 77-92.
81
Strozzi, T., Wegmüller, U., Wiesmann, A. and Werner, C. 2003. Validation of the X-SAR
SRTM DEm for ERS and JERS SAR geocoding and 2-pass interferometry in alpine
regions Proceedings IGARSS'03 (Toulouse, France).
Strozzi, T., Kääb, A. and Frauenfelder, R. 2004. Detecting and quantifying mountain
permafrost creep from in situ inventory, space-borne radar interferometry and airborne
digital photogrammetry. International Journal of Remote Sensing 25:15, 2919-2931.
Stötter, J., Wastl, M., Caseldine, C. and Häberle, T. 1999. Holocene paleoclimatic
reconstruction in northern Iceland: approaches and results. Quaternary Science
Reviews 18:457-474.
Summerfield, M.A. 1991. Global geomorphology: an introduction to the study of landforms.
Longman, Edinburgh, p. 537.
Sun, G., Ranson, K.J., Kharuk, V.I. and Kovacs, K. 2003. Validation of surface height from
Shuttle Radar Topography Mission using Shuttle laser altimeter. Remote Sensing of
Environment 88:4, 401-411.
Toutin, T. 2001. DEM generation from new VIR Sensors: Ikonos, ASTER and Landsat 7
IGARSS'01 July 9-13, 2001 (Sidney, Australia).
Toutin, T. 2002. DEM from stereo Landsat 7 ETM (super +) data over high relief areas.
International Journal of Remote Sensing 23:10, 2133.
Tønsberg, O.M. 2003. Kartlegging av bredynamikken på en fremrykkende bre -Nigardsbreen,
Sør-Norge - vha GPS og fotogrammetri (in Norwegian). Departement of Physical
Geography, University of Oslo. M.Sc.-thesis
Ulaby, F.T., Moore, R.K. and Fung, A.K. 1982. Microwave Remote Sensing: Active and
Passive Radar remote sensing and surface scattering and emission theory. 2.
Addison-Wesley Pub. Company.
Unwin, B. and Wingham, D. 1997. Topography and dynamics of Austfonna, Nordaustlandet,
Svalbard, from SAR interferometry. Annals of Glaciology 24:403-408.
van Everdingen, R. (ed.) 1998. Multi-language glossary of permafrost and related ground-ice
terms (revised May 2005). Boulder: National Snow and Ice Data Center/World Data
Center for Glaciology.
Wagner, S. 1992. Creep of alpine permafrost, investigated on the Murtel rock glacier.
Permafrost and Periglacial Processes 3:157-162.
Wang, Z. and Li, S. 1999. Detection of winter frost heaving of the active layer of Arctic
permafrost using SAR differential interferograms IGARSS'99 June 28-July 2, 1999
(Hamburg, Germany). 4, 1946-1948.
Washburn, A.L. 1979. Geocryology A survey of periglacial processes and environments.
Edward Arnold, London, p. 406.
Weertman, J. 1957. On the sliding of glaciers. Journal of Glaciology 3:21, 33-38.
Wegmüller, U., Werner, C.L. and Rosen, P.A. 1993. Derivation of terrain slope from SAR
interferometric phase gradient Proceedings of the second ERS-1 symposium
(Hamburg, Germany). 711-715. European Space Agency, Paris.
Wehr, A. and Lohr, U. 1999. Airborne laser scanning - An introduction and overview. ISPRS
Journal of Photogrammetry & Remote Sensing 54:2-3, 68-82.
Weydahl, D.J. 2001. Analysis of ERS tandem SAR coherence from glaciers, valleys, and
Fjord Ice on Svalbard. IEEE Transactions on Geoscience and Remote Sensing 39:9,
2029-2039.
Whalley, W.B. 1974. Origin of rock glaciers. Journal of Glaciology 13:68, 323-324.
Williams, P.J. and Smith, M.W. 1989. The Frozen Earth: Fundamentals of Geocryology.
Cambridge University press, Cambridge, p. 306.
Young, A. and Saunders, I. 1986. Rates of surface processes and denuation. In: Abrahams,
A.D. (ed.) Hillslope processees. Allen & Unwin, Bostoun, 3-27 p.
82
Zebker, H.A., Werner, C.L., Rosen, P.A. and Hensley, S. 1994. Accuracy of topographic
maps derived from ERS-1 interferometric data. IEEE Transactions on Geoscience and
Remote Sensing 32:823-836.
Zomer, R., Ustin, S. and Ives , J. 2002. Using satellite remote sensing for DEM extraction in
complex mountainous terrain: landscape analysis of the Makalu Barun National Park
of eastern Nepal. International Journal of Remote Sensing 23:1, 125-143.
Ødegård, R.S. and Sollid, J.L. 1993. Coastal cliff temperatures related to the potential for
cryogenic weathering processes, western Spitsbergen, Svalbard. Polar Research 12:1,
95-106.
Ødegård, R.S., Etzelmüller, B., Vatne, G. and Sollid, J.L. 1995. Near-surface spring
temperatures in an Arctic coastal cliff: Possible implications of rock breakdown. In:
Slaymaker, O. (ed.) Steepland geomorphology. Wiley, Chichester, 89-102 p.
Østrem, G., Liestøl, O. and Wold, B. 1976. Glaciological investigations at Nigardsbreen,
Norway. Norsk Geografisk Tidskrift - Norwegian Journal of Geography 30:1, 187209.
Østrem, G., Dale Selvig, K. and Tandberg, K. 1988. Atlas over breer i Sør-Norge - Atlas of
Glaciers in South Norway. Meddelelse nr. 61 fra Hydrologisk Avdeling. Norges
Vassdrags- og Energiverk, Oslo, p. 248.
83
Part II
Papers
ARTICLE IN PRESS
Geomorphology xx (2006) xxx – xxx
Surface displacements and surface age estimates for creeping
slope landforms in Northern and Eastern Iceland using
digital photogrammetry
B. Wangensteen a,⁎, Á. Guðmundsson b , T. Eiken a , A. Kääb c ,
H. Farbrot a , B. Etzelmüller a
a
c
Department of Geosciences, University of Oslo, P.O. Box 1047 Blindern, NO-0316 Oslo, Norway
b
Jarðfræðistofán-JFS Geological services, Raudðagerði 31, IS-108 Reykjavik, Iceland
Department of Geography, University of Zürich-Irchel, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland
Accepted 27 January 2006
Abstract
In this study three areas of different active, coarse-debris, slope processes are investigated in Northern and Eastern Iceland.
Surface displacement of some glacier-derived rock glaciers and a debris-covered glacier in the Hólar area, a moving debris
accumulation close to Siglufjörður and a debris layer in Seyðisfjörður are measured. The displacement fields are obtained by crosscorrelation matching of multi-temporal orthophotos. Orthophotos are generated using a Z/I-Imaging digital photogrammetric
workstation and various series of air photos from 1964 to 1994. Cross-correlation matching is done with the CIAS software. The
results are analyzed and used for rough surface age estimates. In addition, type and cause of movement are discussed. The velocity
of the debris-covered glacier and the rock glaciers in the Hólar area averages from 0.14 to 0.67m a− 1. The debris accumulation at
Almenningsnöf close to Siglufjörður shows an average displacement of 0.19m a− 1 with a maximum value of 0.84m a− 1. The
displacements at Almenningsnöf agree well with displacements surveyed by GPS by the Icelandic road authorities. The measured
velocities in Seyðisfjörður, although using air photos taken 30 years apart, turned out not to be significant, but the homogenous
direction of the displacement vectors suggests that the debris is currently creeping. Based on the surface age results all the
landforms in Hólar are suggested to have developed during the late-Holocene cooling period, with ages from around 1500 and
3000 to around 5000 years for the different landforms. These surface age estimates coincide with data from moraine datings nearby
and Holocene climatic development in the North Atlantic region.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Surface displacement; Digital photogrammetry; Rock glaciers; Ice-cored moraines; Surface age estimates; Iceland
1. Introduction
⁎ Corresponding author. Tel.: +47 22 85 73 18; fax: +47 22 85 42 15.
E-mail address: bjorn.wangensteen@geo.uio.no
(B. Wangensteen).
0169-555X/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2006.01.034
In Tröllaskagi, the high mountain areas of northern
Iceland, small cirque glaciers, ice-cored moraines and
talus-derived slope accumulation landforms (rock
glaciers) are frequent in the periglacial zone above
800 m a.s.l., and comprise an important part of the
ARTICLE IN PRESS
2
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Holocene sediment flux system of the landscape. Martin
and Whalley (1987) and Whalley and Martin (1994)
have described and analyzed some of the features,
showing in those cases that they consisted of sedimentary ice, covered by coarse debris, creeping some tens of
centimetres per year. Those authors defined all these
landforms as rock glaciers. Furthermore, they stated that
these areas lack permafrost conditions and that the
landforms are mostly related to what they assumed to be
temperate glacier ice. Thus, they argued against the
concept of Barsch and Haeberli (Barsch, 1992; Barsch,
1996; Haeberli, 2000) that rock glaciers are normally an
expression of creeping permafrost. Recently, these
features were systematically mapped by Guðmundsson
(2000), showing a high abundance of active, rock
glacier-like forms in certain parts of northern and eastern
Iceland. Such slope processes are linked to climatic
conditions, and knowledge of these processes can
therefore aid interpretations of climatic change through
the Holocene and its impact in these environments.
Large coarse-grained slope accumulation landforms
are also common throughout northern and eastern
Iceland at lower altitudes along the major valleys and
coastal slopes. Many of these forms display characteristics of movement and creep as transversal ridges and
furrows. Roads and buildings situated on or below such
landforms experience damage from soil creep or debris
flows. At lower altitudes these landforms are prominent
features in the landscape. Þorarinsson et al. (1959)
describe them as large landslides, probably formed early
after deglaciation due to valley wall stress release. The
explanation is commonly accepted among geoscientists
in Iceland. However, recently Guðmundsson (2000)
proposed the hypothesis that these slope features may be
related to a periglacial climate in association with
permafrost conditions during their development, and
classified them as “fossil rock glaciers”.
With respect to the systems defined by Caine (1974)
in models of geomorphic activity in alpine environments, the high mountain areas (Hóladalur) in this study
fall within the slope system. In addition Caine (1974)
also defines a stream channel system for such environments. The slope system is subdivided into input,
transfer and storage, and the processes investigated in
this study are within the transfer or mass wasting
subsystem. For the two other sites the investigations also
concentrate on mass wasting in the slope system, while
another classification is needed especially at Almenningsnöf, since no stream channel system is present but
rather a geomorphic system involving coastal processes.
The objective of the study is to map the displacement
field of selected creeping slope landforms and in thick
sediment-covered slopes in northern and eastern Iceland
by means of digital photogrammetry. The main goal is to
quantify the magnitude and investigate the nature of the
displacements, and to use these results for surface age
estimates of the landforms. At present, there are very
limited data available on the nature and rate of slow
slope movements in Iceland. This results in a limited
basis for interpretations of the types of processes
present, and for estimating the current and Holocene
average denudation rates. In this study we have selected
three sites, each of them representing different kinds of
slope landforms and processes. The first site near Hólar
represents a series of active slope landforms at an
altitude of 850 to 1000 m a.s.l. The second site,
Almenningsnöf close to Siglufjörður, is a huge debris
accumulation covering altitudes between sea level and
260m a.s.l. The third site is close to the village of
Seyðisfjörður, covering a debris-mantled slope of up to
70m thick sediments, which creeps slowly (Fig. 1).
2. Nomenclature used in this paper
The high-mountain landforms investigated in this
paper are highly debated, both in Iceland and internationally. According to the concept of Barsch (1992,
1996) and Haeberli (2000) most of the debris bodies
defined as rock glaciers by Whalley and Martin (1994),
would be creeping ice-cored moraines since glacier or
sedimentary ice probably form the core of the debris
bodies. Barsch and Haeberli use the term rockglacier (in
one word) solely for perennially frozen debris supersaturated with interstitial ice and ice lenses that endure
creep due to deformation of ice. This includes the
cohesive flow of talus rockglaciers and debris rockglaciers, but excludes features originating from sedimentary glacier ice. Humlum (1982) distinguish
between talus- and glacier-derived rock glaciers (in
two words), independent of ice origin, but with the
precondition of permafrost existence. In our paper we
follow the discussion by Humlum, defining the slope
features in Hóladalur and Fremri-Grjótárdalur as rock
glaciers if they are more or less totally covered by
debris. One of the landforms is clearly not totally debriscovered and the term debris-covered glacier is therefore
used. By using the term rock glacier, we assume that the
landforms are situated in a permafrost area. The
assumption of permafrost will be discussed later.
The large debris bodies at lower altitudes are defined
as large landslides in the scientific literature, indicating a
singular or episodic formation (Þorarinsson et al., 1959).
However, many of them are still in a state of creep
according to measurements done by the Icelandic road
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
3
Fig. 1. The location of the three investigated sites; Hóladalur and Fremri-Grjótárdalur (1), Almenningsnöf (2) and Seyðisfjörður (3).
authorities (Sæmundsson, 2004). This paper does not
aim to discuss the initial conditions of the landform
development, but aims to assess the present movement
pattern in the light of displacement rates, and possible
surface ages inferred from that. In this paper we
therefore use the descriptive term debris accumulation
for these landforms, inferring that in an earlier
periglacial environment, creep, in addition to slide
events, could have contributed to the present-day
landform.
3. Setting
3.1. Hóladalur
Hóladalur and Fremri-Grjótárdalur are located close
to Hólar on the central part of the Tröllaskagi peninsula
(65°40′N, 19°W) (Fig. 1). The Tröllaskagi peninsula is
situated in northern Iceland between Skagafjörður in the
west and Eyafjörður in the east and made up of land
ranging from 600 to 1400 m in altitude having basalts of
Upper Tertiary age as its main bedrock component
(Þordarson and Hoskuldsson, 2002). Deep glacial
valleys and fjords cut into the mountainous peninsula.
At least 165 rock glacier-like forms and ice-cored
moraines are found to be active on this 4800 km2
peninsula (Guðmundsson, 2000).
The studied landforms fill the valley ends of
Hóladalur and Fremri-Grjótárdalur (see Fig. 2). A big
(∼ 4km2) debris-covered glacier dominates the upper
valley of Hóladalur (Fig. 3A). Adjacent is a glacierderived rock glacier. In the side valley, FremriGrjótárdalur, a complex of glacier-derived rock glaciers
fills the inner part of the valley (Fig. 3B). All these
landforms are situated between 900 and 1200 m a.s.l.
within two cirques, surrounded by mountains and
mountain plateaus reaching 1300 m a.s.l. The age of
the basalt in this area is about 7 million years (Þordarson
and Hoskuldsson, 2002).
The debris-covered glacier (nr. 1 in Fig. 2) has a
debris layer which covers 2/3 of the glacier area in the
1964 photos, reaching a thickness of several metres. A
debris layer will protect the glacier surface from melting
(Østrem, 1959) and in this case the glacier extends far
below what would have been the case under debris free
conditions. The debris cover consists of unsorted enand supraglacially transported angular material that
probably has reached the glacier as rock falls from the
ARTICLE IN PRESS
4
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 2. Overview of the investigated landforms in Hóladalur and Fremri-Grjótárdalur. The numbers refer to the areas in Table 3.
headwall of the cirque. The debris-covered glacier has
an area of about 4.3 km2 of which 3.4 km2 is covered by
the orthophotos used.
The rock glacier adjacent to the debris-covered
glacier (nr. 2 in Fig. 2) is rather small and hummocky,
and believed to originate from a partly melted ice-cored
Fig. 3. A) Debris-covered glacier in Hóladalur (photo: Ágúst Guðmundsson), B) active rock glacier complex in Fremri-Grjótárdalur (photo: Herman
Farbrot) and C) debris accumulation at Almenningsnöf (air photo: Landmælingar Islands).
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
moraine. The size is 0.29 km2 and it is situated between
900 and 1000m a.s.l.
The rock glacier complex in the neighbouring cirque
in Fremri-Grjótárdalur (Fig. 3B) only shows glacier ice
in small patches along the headwall. However, the
whole landform is probably underlain by ice of
unknown origin, indicated by DC resistivity measurements (Farbrot, unpublished data). These landforms
also depict what may be several rock glacier generations, and like the neighbouring debris-covered glacier,
rock falls from the headwalls of the cirque are the
probable main source of material input. The eight
identified landforms in this rock glacier complex are
numbered 3 to 10 in Fig. 2 and vary in area from 0.031
to 0.435 km2, with altitudes ranging from 900 to 1160m
a.s.l. What appears to be a small talus-derived rock
glacier located centrally below the headwall of the
same cirque is numbered 11 in Fig. 2 and has an area of
0.047 km2. All the landforms described exist due to
relative high weathering rates under periglacial conditions. The basalt bedrock is very susceptible to both
frost weathering and chemical weathering (see e.g.
Beylich, 2000).
There are no temperature measurements in Hóladalur
and Fremri-Grjótárdalur, but MAAT is probably around
− 3.0 °C at the front (900 m a.s.l.) of the investigated
landforms at this site. This temperature is extrapolated
from a MAAT of 2.4 °C in Hólar at 140m a.s.l. (The
Icelandic Meteorological Office), some 4–6km away,
using the temperature gradient of − 0.71 °C/100 m
suggested by Gylfadóttir (2003).
3.2. Almenningsnöf
The second location is the moving debris accumulation at Almenningsnöf between Kvígildi and Skriðnavik on the coast road to Siglufjörður at the northern tip
of the Tröllaskagi peninsula (66°10′N, 19°W) (Fig. 1).
The bedrock of this area is the same basalt of Upper
Tertiary age that is found in Hólar, but slightly older
with an age of about 12 million years (Þordarson and
Hoskuldsson, 2002). However, the relief of the northern
tip of the peninsula is more alpine than at Hólar.
Geodetic displacement measurements along the road
that crosses the moving debris accumulation have been
undertaken since 1977 by the Icelandic road authorities.
They revealed yearly displacements of up to more than
1 m (Sæmundsson, 2004). These measurements were
initiated due to observed damage to the road, and are
used as a reference for the displacements derived from
photogrammetry in this study. The debris accumulation
is situated below north- and west-facing rock walls,
5
between sea level and 260m a.s.l. (Fig. 3C). It comprises
a coast of 2 km in length and has an area of 1.9km2. The
greatest changes in elevation are found close to the sea
with the 100m a.s.l. contour line 200–450m inland. The
distance from the headwall to the sea is 1.7km (Fig. 4).
The origin of the debris accumulation is uncertain, and
different hypotheses exist. Sæmundsson (2004) treats
the landform as a result of repetitive rock slides
connected to the isostatic rebound after the last
glaciation. In contrast, Guðmundsson (2000) argues
that the landform is a result of local glaciation and
permafrost conditions after or during the last glaciation
of Iceland, and also implies that there might be buried
ice within the debris accumulation. Guðmundsson
(2000) reports on debris thicknesses in the range 15 to
65 m from core drillings along the road, with an average
value of 42 m. Sæmundsson (2004) assumes a mean
debris thickness of 50 m and a total volume of
110,000,000 m3 for the whole accumulation.
3.3. Seyðisfjörður
The third location is the debris-mantled slope above
the Seyðisfjörður community (Fig. 1). Here, the layer of
moving debris does not show any specific landform
connected to surface movement in contrast to the two
other investigated areas. Heavy rainfall from time to
time triggers debris flows from this slope, threatening
the infrastructure of the village. The investigated area
covers around 1 km2 of the northeast-heading slope
above the Seyðisfjörður community and ranges from 30
to 340 m in altitude. Between 80 and 120 m a.s.l. there is
a bedrock threshold with cliffs. The upper part of the
slope is also inclined with a gradient around 6%.
Elsewhere, the slope is gentler with a gradient between 1
and 3%. Drilling has revealed debris thicknesses of 16–
71 m (Guðmundsson et al., 2003). The drillings are all
done above the steep bedrock threshold. The bedrock is
exposed in steeper areas. The mountains around the
glacially deeply eroded Seyðisfjörður reach more than
1000 m a.s.l. The bedrock in Seyðisfjörður is also basalt
of Upper Tertiary age (Þordarson and Hoskuldsson,
2002).
4. Technique
Cross-correlation matching has been used for
mapping glacier velocities in satellite imagery since
the early 1990s (Bindschadler and Scambos, 1991) and
in aerial orthophotos since 1995 (Rolstad, 1995). Aerial
photogrammetry has been used as a tool for velocity and
geometry change measurements of glaciers, landslides
ARTICLE IN PRESS
6
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 4. A geomorphologic sketch of the investigated area at Almenningsnöf. The numbers refer to the areas in Table 3.
and creeping permafrost landforms like rock glaciers for
several decades, and since the advent of digital
photogrammetry cross-correlation matching of orthophotos has also been widely used (Baltsavias, 1996;
Kääb and Vollmer, 2000; Kaufmann and Ladstädter,
2003; Delacourt et al., 2004; Roer et al., 2005).
In this study orthophotos are generated using a Z/IImaging digital photogrammetric workstation (DPW)
with aerial photography obtained from Landmælingar
Islands (National Land Survey of Iceland) (Table 1).
Orthophotos are aerial photos transformed to orthogonal
projection (i.e. map projection) by the use of air photos
and a digital terrain model (DTM) within a DPW
(Kasser and Egels, 2002). Since we had no DTM
available, the first step in creating the orthophotos was
to generate stereo models from the overlapping air
photos of each location and year. Ground control points
(GCP) used for the absolute orientation of the stereo
model were collected during a field campaign in August
2003 using differential GPS. To increase the accuracy
between the stereo models of different years homologous tie points and the same GCPs are used for all
Table 1
Data on the aerial photos used, and the generated orthophotos and DTMs
Location
Air photo date
Air photo scale
(approximately)
Air photo
resolution (m)
Orthophoto
resolution (m)
DTM
resolution (m)
Hóladalur
Hóladalur
Almenningsnöf
Almenningsnöf
Almenningsnöf
Seyðisfjörður
Seyðisfjörður
06
07
13
06
07
09
09
1:29,000
1:29,000
1:16,000
1:35,000
1:35,000
1:13,500
1:15,500
0.37
0.37
0.21
0.45
0.44
0.17
0.24
0.5
0.5
0.5
0.5
0.5
0.3
0.3
5 and 20
5 and 20
9
9
9
5
5
Aug 1985
Aug 1994
Aug 1977
Aug 1985
Aug 1994
Sep 1964
Aug 1994
At Hòlar 20 m resolution DTMs were used for the debris-covered glacier, while 5m DTMs were used for the rock glacier complex.
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
models for the same area. Having a stereo model it is
possible to automatically generate a DTM. Z/I-Imaging
uses the Match-T algorithm for this purpose (Krzystek,
1991). An orthophoto is then generated for each year at
each location using orientation parameters and the
corresponding DTM. The orthophotos are generated
with 0.3 and 0.5 m resolution with DTMs having a
resolution between 5 and 20m (see Table 1). Coarser
models are used at the locations with small variability in
topography. The residuals from the relative orientation,
including the co-registering of the orthophotos, are
between 0.36 and 0.54m.
The cross-correlation matching of orthophotos is
done with the CIAS software (Kääb and Vollmer, 2000).
This software matches homologous points in two geoand co-referenced orthophotos of the same area taken at
different times. The points are either selected manually
in the orthophoto of time 1 or as points of a regular grid
with a specified grid size in the orthophoto of time 1. A
reference block, a small window, is extracted from the
orthophoto of time 1 around each selected point and the
homolog of this small window is searched for in a larger
test area around the corresponding coordinate in the
orthophoto of time 2. A cross-correlation factor is
calculated for each possible location of the reference
block within this test area. The location that yields the
highest correlation factor is taken to be the new position
in orthophoto 2 of the point selected in orthophoto 1. For
a thorough description see Kääb and Vollmer (2000). If
a displacement has taken place during the time interval
between the two acquisitions, this displacement is
measured as the horizontal distance between the two
homologous points (i.e. this method does not measure
the vertical displacement). A reference block size of
15 × 15 pixels was used for all locations. This size seems
to be sufficient to contain enough information for
matching 0.3 and 0.5 m resolution photos of the coarsegrained boulders found on the landforms mapped in this
study. Boulders ranging in diameter from 1 m to a few
metres were used for the manually selected points. The
size of the test area was 50 × 50 pixels for the slow
movements and 100 × 100 pixels for the faster ones. The
test area has to be large enough to detect a displacement
of the expected magnitude, and the size was determined
by some initial testing. According to Kääb and Vollmer
(2000) the accuracy of the method is in the order of one
pixel and at least at the same level as results from
conventional photogrammetry. Although originally used
for deformation mapping of rockglaciers by the use of
aerial orthophotos, this method has also been used to
map glacier movement using ASTER data (Kääb,
2002).
7
After a sufficient number of points have been
matched by cross-correlation, it is possible to perform
filtering on the resulting data. Displacement vectors
with low cross-correlation factors or unnatural direction
can easily be filtered out. A cross-correlation threshold
of 0.5 was used for all locations together with a
directional filter, filtering any vector deviating greatly
from the assumed flow direction. The filtered data were
then imported in to the GIS ArcView and some manual
removal of points was also undertaken. In the areas
where manual point picking was used the filtering and
manual edit removed about 25% of the original matched
points, while being closer to 50% for the areas where a
regular grid was used to define the points.
By integrating streamlines through time and the
measured displacements, surface age estimates can be
obtained for the landforms. This is based on the
assumption that the present velocities can be taken as
a constant velocity for the whole existence of the
landform and thereby give the time it takes for a surface
particle to travel the whole length of the debris cover.
This is of course a crude surface age estimate, but
nevertheless it can give some valuable information on
the history of the landforms. The streamlines are
generated using the technique discussed by Kääb et al.
(1998). The velocity profiles presented are taken from
displacement fields interpolated from the measured
displacements by the use of Kriging. Both interpolation
and velocity profiling are done using ESRI Arc Map.
5. Displacements and streamlines
5.1. Hóladalur
The combination of the georeferencing and crosscorrelation matching accuracy is about 0.5 m for the data
from Hóladalur and Fremri-Grjótárdalur, and for the
period of 9 years between 1985 and 1994 the
displacements must be larger than 0.06m a− 1 in order
to be significant (see Table 2). The displacements of the
debris-covered glacier (Fig. 5) have a pattern of
Table 2
Time interval between the air photo acquisitions, with total and per
year accuracy for the displacement measurements
Location
Time interval
Accuracy
(m)
Accuracy
(cm a− 1)
Hóladalur
Almenningsnöf
Almenningsnöf
Seyðisfjörður
1985–1994
1977–1985
1985–1994
1964–1994
0.5
0.5
0.5
0.3
5.56
6.25
5.56
1.00
ARTICLE IN PRESS
8
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 5. Measured displacements for the debris-covered glacier and the rock glacier in Hóladalur (nr. 1 and 2 in Fig. 2) based on the 1985 and 1994 air
photos.
increasing velocities towards the middle and also
towards the terminus, and range from less than 0.10m
a− 1 to a maximum of 0.84 m a− 1 at the front. This is also
seen from the profiles in Fig. 6. The areas of the greatest
displacement coincide with the areas of steepest slope.
The small rock glacier located west of the debriscovered glacier in Hóladalur (nr. 2 in Fig. 2) only gave
very few matched points with velocities ranging from
around 0.11 m a− 1 to 0.39 m a− 1.
The displacements of the rock glacier complex in
Fremri-Grjótárdalur show a different pattern with the
greatest displacement for the upper lobes, and for form
nr. 7 (Fig. 2) also increasing velocities with elevation
(Figs. 7 and 8). Maximum velocity for the rock glacier
complex is 0.74m a− 1. One can also see that the
different generations outlined in Fig. 2 show differences
in displacement, with the uppermost and youngest forms
moving 2–3 times faster than the lower and older ones.
Landform nr. 3 (Fig. 2) is believed to be a collapsed rock
glacier and has the smallest displacements. The DTMs
and orthophotos show that both the geometry and front
position for all the ten identified landforms have been
stable over the period from 1985 to 1994, considering
the significance level of data and measurements.
The streamlines of the Hóladalur debris-covered
glacier inferred from the 1985–1994 velocities suggest a
total surface development age for the debris cover of
about 4500 to 5000 years, assuming the present velocity
field to be representative over a long time span (Fig. 9).
For the rock glacier complex in Fremri-Grjótárdalur the
surface development ages of the youngest and upper
forms are around 1000 to 1500 years (Fig. 10). For the
talus-derived rock glacier (nr. 11 in Fig. 2) and the three
easternmost forms (8, 9 and 10 in Fig. 2) there are not
enough displacement measurements to establish reliable
streamlines. The velocity of the lowermost and
collapsed form is very low and implies an additional
surface age of at least 1500 years. In the northernmost
part of this form few measurements inhibit any surface
age estimates to be made.
5.2. Almenningsnöf
The velocity measurement accuracy at Almenningsnöf is believed to be around 0.5m, so that any significant
displacements must be larger than 0.06 m a− 1 (Table 3).
The displacement based on orthophotos from 1985 to
1994 from Almenningsnöf depicts the greatest velocities
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
9
Fig. 6. Velocity profiles A–A′ and B–B′ on the debris-covered glacier in Hóladalur. The location of the profiles is shown in Fig. 2.
in the south-western part with an average of 0.57m a− 1
(Fig. 11A). The velocities are around 0.33 m a− 1 in the
mid section and around 0.13 m a− 1 in the upper part of
the body. The part of the body draining northwest has an
average velocity of 0.17m a− 1. At some places the
movements show a discrete rather than a continuous
pattern where the velocity changes rapidly over small
areas (Fig. 12). Here the velocity differs greatly on the
two sides of a marked fissure parallel to the direction of
movement. This type of fissure can be found on both
sides of the fast moving area 1 below the road. The
overall pattern and magnitude of the displacement are
also confirmed by the cross-correlation matching of all
sets of orthophotos from 1977, 1985 and 1994 (Fig. 11A,
B), with the same average velocity of 0.19m a− 1 for the
1977–1985 and 1985–1994 periods (Table 3). For the
four different areas, velocities are more or less the same
for the two periods, except for the fastest moving area in
the south-western part of the body (area 1, Table 3). In
this area the average velocity is 0.38 m a− 1 from 1977 to
1985 and 0.57m a− 1 from 1985 to 1994. There is also a
slight difference for area 2 (Table 3). The velocity
measurements from 1977 to 1985 seem to have a higher
density, with 2422 points matched compared to the 792
points from the 1985 and 1994 orthophotos. The
measurements are also in agreement with the GPS
measurements along the road (Fig. 11).
The streamlines inferred from the 1985 to 1994
velocities predict a travel time of at least 7000 years
(Fig. 13). This is the time it takes for a surface particle to
travel from the upper areas of the landform and down to
the shore, using modern velocities for extrapolation.
Since there is coastal erosion taking place along the
terminus, and the material origin is uncertain, this
surface age estimate based on travel time should be
treated with great care. The 1977–1985 data give a
travel time of at least 5000 years and the trajectories are
much straighter than the ones from the later period (Fig.
13). From Fig. 12 it can also be seen that it was not
possible to start any of the trajectories all the way back
at the headwall, due to few matched points in the 1985–
1994 data set. Extrapolating the trajectories back to the
ARTICLE IN PRESS
10
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 7. Measured displacements for the rock glacier complex in Fremri-Grjótárdalur based on the 1985 and 1994 air photos.
headwall would increase the estimated travel time 60–
70%, resulting in surface age estimates around 8000 and
12,000 for the 1977–1985 and 1985–1994 data sets.
The accuracy of an automatically generated DTM
from air photos taken from an altitude of 5500 m (1985
and 1994 photos) is 1–2.5 m (0.2–0.4‰ of the flying
height), hence, differencing two DTMs adds up the error
to 2–5 m. Taking into account this degree of accuracy
the difference between the DTMs does not reveal any
significant changes in geometry for any of the
investigated areas. Only an area some 200 m north of
area 1 at Almenningsnöf, where road works have
lowered the surface with more than 15 m, is taken to be a
significant change.
5.3. Seyðisfjörður
Velocity measurements for Seyðisfjörður based on
photos from 1964 and 1994 show velocities ranging
between 0 and 0.13 m a − 1 , reaching the highest
velocities in the two steeper areas (Fig. 14). In the
more gently sloping areas in between, velocities are
lower, averaging around 0.01–0.02m a− 1. The overall
accuracy is about 0.3m corresponding to a displacement
of 0.01 m a− 1 over the 30-year period (Table 3). This
implies that the measured displacements in Seyðisfjörður in general are not significant given the 0.01 m a− 1
accuracy.
6. Discussion
6.1. Accuracy and technique evaluation
Kääb and Vollmer (2000), who developed the CIAS
software, estimate an average deviation originating from
the cross-correlation method, relative orientation and the
DTM to be about one pixel. In our case this means 0.5m
at Hólar and Almenningsnöf and 0.3m at Seyðisfjörður.
Given this orthophoto resolution and the time span
between the acquisitions of the air photos, the one pixel
error therefore results in a 5.6 cm a− 1 error for the 1985–
1994 data from Hólar and Almenningsnöf, a 6.3 cm a− 1
error for the 1977–85 data from Almenningsnöf and a
1.0 cm a − 1 error for the 1964–1994 data from
Seyðisfjörður (Table 2). Looking at the average
velocities in Table 3 this means that all measured
displacements can be taken to be significant, except for
the measurements from area 3 at Almenningsnöf and at
Seyðisfjörður where the measured values are just twice
the expected error. But for all these areas the direction of
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
11
Fig. 8. Velocity profiles C–C′ and D–D′ from the rock glacier complex in Fremri-Grjótárdalur. The location of the profile is marked in Fig. 2.
the displacement shows a clear trend perpendicular to
the contour lines. This is in contrast to areas outside the
accumulation where the direction of the displacements
has a more chaotic and random appearance. Velocities
with a yearly displacement of more than 0.06m at Hólar
and Almenningsnöf, and 0.01 m at Seyðisfjörður are
therefore included in the result. At Seyðisfjörður the
average displacement is around the accuracy level, but
there are still parts of the areas moving faster with
significant displacements and these are included in the
results.
Comparing the accuracy and applicability of crosscorrelation matching of orthophotos with other methods
like differential GPS surveying, satellite SAR interferometry and permanent scatter technique, the main
advantage is the combination of accuracy and density
of measurements. Differential GPS will achieve a higher
accuracy for each measured point, down to less than a 1
cm per measurement and around 1.5 cm when measuring displacements (Berthling et al., 1998). But it is very
time consuming and costly to measure displacements
with the same density as achieved by cross-correlation
matching of orthophotos. Using satellite SAR interferometry one can measure displacements down to 1.5 mm
for vertical movements and 4mm for horizontal movements (Goldstein et al., 1993), but the spatial resolution
is only around 20 m. For satellite SAR interferometry
there is also the aspect of decorrelation between
acquisitions due to temporal effects (Weydahl, 2001),
and the availability of scenes with surface motion
towards the satellite sensor (i.e. orbit). The permanent
scatter technique in SAR interferometry faces fewer
problems concerning decorrelation. Only objects with
strong radar back scatter can be used (e.g. constructions,
visible bedrock etc.), and for the areas used in this study
such strong reflectors are scarce. The technique also
requires long series of data (Ferretti et al., 2000; Ferretti
et al., 2001).
On Almenningsnöf the difference in travel times
from the 1977–1985 and the 1985–1994 data set is the
most striking, having streamlines giving 5000 years for
the first period and 7000 for the second (Fig. 13). Both
displacement data sets are estimated to have a similar
accuracy. If one investigates the data closely it is
apparent that the streamlines from the second period
are much more curved and undulating. They also show
low displacements in areas where the velocity vectors
differ in direction. The method uses the 10 closest
ARTICLE IN PRESS
12
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 9. Streamlines for the debris-covered glacier in Hóladalur (nr. 1 in Fig. 2) based on the displacement field from the 1985–1994 data.
Fig. 10. Streamlines for the rock glacier complex in Fremri-Grjótárdalur based on the displacement field from the 1985–1994 data.
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Table 3
Measured velocities at the three locations Hóladalur, Almenningsnöf
and Seyðisfjörður
Location
Time
period
Average
velocity
(m a− 1)
Maximum
velocity
(m a− 1)
Hóladalur 1
Hóladalur 2
Hóladalur 3
Hóladalur 4
Hóladalur 5
Hóladalur 6
Hóladalur 7
Hóladalur 8
Hóladalur 9
Hóladalur 10
Hóladalur 11
Almenningsnöf
Almenningsnöf
Almenningsnöf 1
Almenningsnöf 1
Almenningsnöf 2
Almenningsnöf 2
Almenningsnöf 3
Almenningsnöf 3
Almenningsnöf 4
Almenningsnöf 4
Seyðisfjörður
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1985–1994
1977–1985
1985–1994
1977–1985
1985–1994
1977–1985
1985–1994
1977–1985
1985–1994
1977–1985
1985–1994
1964–1994
0.37
0.19
0.14
0.33
0.27
0.46
0.54
0.67
0.67
0.58
0.57
0.19
0.19
0.38
0.58
0.29
0.41
0.15
0.13
0.20
0.17
0.02
0.84
0.39
0.26
0.54
0.44
0.66
0.74
0.72
0.70
0.74
0.89
0.67
0.84
0.62
0.84
0.67
0.80
0.39
0.38
0.51
0.52
0.13
velocity measurements when interpolating. So the
reason for the difference in the estimated travel times
is most probably due to differences in density and
homogeneity of matched velocity points. The reason
for this may be attributed to differences in the data
used. As seen from Table 1 the air photos from 1977
have an original resolution of 0.21 m while the photos
from 1984 and 1994 have a resolution of 0.44 and
0.45 m respectively. All the orthophotos were generated
with a resolution of 0.50 m, and both periods include
the 0.44m resolution air photos from 1985. However,
the better contrast in the orthophotos from 1977
increases the number of points that are possible to
match, and therefore also the density of the resulting
velocity measurements and the quality of the surface
age estimates. Thus, images with dense velocity
measurements should be preferred for surface age
estimates using streamlines. This means that one
should emphasise the travel times from the first period
rather than the last and that the travel time is around
5000 years (i.e. 8000 years interpolating the streamline
back to the headwall). It is also important to note that
this is the time it takes for a surface particle to travel
from the headwall to the shore. Since the origin of the
landform is not known it is difficult to apply this as a
surface age estimate for the landform.
13
6.2. Flow regime
There is very little work done on surface displacement of debris-covered glaciers and rock glaciers in this
area. Martin and Whalley (1987) report of a 2-m
displacement between 1977 and 1985 of the Nautardálur
rock glacier, a debris-covered glacier located some
30 km to the southeast of Hóladalur on the same
peninsula. This corresponds to a yearly displacement
of 0.25 m. From lichenometric studies Hamilton and
Whalley (1995) infer an age of 200 years for the debris
cover of this glacier.
The flow regime of the debris-covered glacier in
Hóladalur (Figs. 5 and 6) seems to be dominated by
glacier flow as the main component of movement.
Here, the velocity increases towards the middle of the
glacier and also downslope. The velocity increase near
the front is more difficult to explain by glacial
dynamics. It is found in an area where the surface
slope is steeper than for the rest of the glacier, but this
area is probably considerably thinner (approximate
thickness of 20–40 m). Hence, an acceleration of a
glacier at this point must include other processes.
Another explanation may be related to periglacial
processes, as an increase in slope would affect any
creep processes taking place within the debris cover
itself. The surface pattern of the debris cover with
furrows and ridges transverse to the flow direction also
implies that there is additional creep taking place in the
debris cover. However, based on our data it is not
possible to be certain on this issue.
The velocity of the adjacent rock glacier in Hóladalur
(Fig. 5) has a more chaotic appearance than the debriscovered glacier and no clear directional trend in the
movements, prohibiting conclusions about flow regime.
The hummocky surface of this landform is most likely
the result of thermokarst and appears to be formed by a
partly melted rock glacier.
The velocity measurements of the rock glacier
complex in Fremri-Grjótárdalur (Figs. 7 and 8) reveal
different velocities for the different landforms identified.
The upper rock glaciers have a greater velocity than the
lower ones. This strengthens the idea of several
generations of landforms overrunning each other.
Some of the larger rock glaciers also experience an
increase in velocity with altitude. However, there does
not seem to be a clear connection between velocity and
slope. The surface of the rock glacier complex also
shows signs of creep-like furrows and ridges, and the
forms are more pronounced than on the debris-covered
glacier. This might be caused by deforming glacier ice or
some additional surface creep. Based solely on the
ARTICLE IN PRESS
14
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 11. A) Measured displacements of the debris accumulation by Almenningsnöf based on the air photos from 1985 and 1994. GPS measurements
undertaken by Vegagerðin (Icelandic road authorities) are shown for comparison. B) Measured displacements of the debris body by Almenningsnöf
based on the air photos from 1977 and 1985.
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
15
Fig. 12. A) Measured displacements on both sides of a fissure north in area 1 (Fig. 4). The orthophoto is used as background. There is a marked
difference in the velocity on the two sides. B) A photo of the same fissure (photo: Trond Eiken).
motion pattern it is difficult to say anything on the
relative importance of the two causes of flow.
The movement on Almenningsnöf is relatively
uniform in areas 3 and 4 (Fig. 11), with average
velocities of 0.13 to 0.19m a− 1 for both investigated
periods. These areas do not have any smaller area that
differs greatly in velocity from the rest, except from a
small speed up close to the shore. Along both sides of
the debris accumulation there are distinct ridges that
indicate a surface lowering and differential movement of
the accumulation and the surrounding terrain. This
homogenous movement could be due to movement
along a sliding plane or deformation within the debris
itself; however, it is difficult to interpret the cause of
movement from the displacements and topography
alone. An exception of that is the fast moving area 1
ARTICLE IN PRESS
16
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Fig. 13. Streamlines for the site at Almenningsnöf based on the displacement field from A) the 1985–1994 data and B) the 1977–1985 data.
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
17
Fig. 14. Measured displacements for the debris layer by Seyðisfjörður based on the air photos from 1964 and 1994.
(Fig. 4). High and uniform velocities in this area (Fig.
11) indicate that this part of the debris accumulation is
moving independently from the rest. The homogenous
displacement indicates some kind of slide movement
along a horizon within the mass. Since the surface is
quite planar and does not show any vertical movement
one can assume that the slip surface also is planar. This
indicates a translational slide (Cruden and Varnes, 1996;
Dikau et al., 1996), and considering the high debris
thickness (Guðmundsson, 2000) we assume this area
being a slow moving slab slide (Dikau et al., 1996). The
assumption of a translational slide is also strengthened
by the fact that the movements are not constant. Instead,
they increase in wet periods and almost cease in dry
periods just like the large annual variability seen in
Sæmundsson (2004). Velocities increase more than
ARTICLE IN PRESS
18
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
fivefold during wet years for some of the GPS points
measured by the Icelandic road authorities in area 1. The
movement of such slides usually takes place along an
abrupt change in rock type or along a permeable/
impermeable soil junction (Dikau et al., 1996). Area 2
shows scarps which are typically seen at the back of
transitional slides (Dikau et al., 1996). The increased
velocity in area 2 can be explained both by the slope
angle and by the stress relieving effect of the fast
moving area in front.
The velocities measured in Seyðisfjörður (Fig. 14)
are much lower than the ones measured at the two other
locations and also close to or at the level of accuracy for
the method used. From the measured velocities one can
see that velocity increases with slope angle and that the
greatest displacements are found close to the bedrock
threshold 80 to 120 m a.s.l and also in the steeper upper
part of the investigated area. There are few points
matched in the relative flat area in between these two
steep areas due to few blocks of a desirable size. The
slow displacement rates imply that the movement occurs
at shallow depth. One should not elaborate more on
these results since they do not show the desired
significance.
6.3. Surface age estimations and preliminary implications on Holocene landscape development
For the debris-covered glacier in Hóladalur the
trajectories suggest a surface age between 4500 and
5000 years for the debris-covered glacier, while the
more active landforms in the rock glacier complex in
Fremri-Grjótárdalur have a suggested surface age of
1000 to 1500 years. The lowermost and slow moving
landform of the complex has an additional surface age of
at least 1500 years. This means that landforms in both
cirques have been initiated between 3000 and 5000
years ago. This agrees well with the onset of Holocene
cooling 4000 years BP found by ice core drilling on
Greenland (Dahl-Jensen et al., 1998) and the onset of
neo-glaciation of western Norway around 5300 years
BP (Nesje, 1992).
The assumption of constant velocity during the last
3000 to 5000 years is of course crude. The velocities
may have been greater during the Little Ice Age due to
increased glacier thickness and probably lower during
the early evolution of the glaciers in this area, while the
colder temperatures during the Little Ice Age would
have had the opposite effect on the debris creep. Both of
these facts contribute to the uncertainty of the surface
age. From lichenometric studies Hamilton and Whalley
(1995) report on an age of 200 years for the debris-
covered glacier in Nautardálur, some 30 km to the
southeast. However, combining present velocity measurements and length profiles for the same glacier they
get an age of 2000 years. The discrepancy is explained
by a greater velocity during the Little Ice Age, which
they believe has been around 2 m a− 1 before slowing
down to the present velocity of 0.25 m a− 1. They
concluded that the initiation of the glacier took place
during the Little Ice Age. In Hóladalur and FremriGrjótárdalur there are no end moraines or glacier
forefields found that support a greater glacier extent
during the Little Ice Age. The surface of the landforms
could have been somewhat steeper during the Little Ice
Age but not to the extent needed to support a velocity
justifying an initiation during the Little Ice Age.
Stötter et al. (1999) report on several glacier
advances during Holocene in the Tröllaskagi area from
moraine dating. In front of Vatnsdalsjökull, situated
some 10km to the northeast of Hóladalur, they found
evidence of advances around 4700 BP and ca. 3200–
3000 BP. In front of Barkárdalsjökull, situated some
10km to the east of Hóladalur, they found evidence of
advances around 2000 BP and 1500 BP. Judging from
the surface age estimates in this study the landforms
could very well have been initiated during the same
periods as the advances for the glaciers in the vicinity.
The debris-covered glacier, which is the oldest landform, may have been initiated during the first reported
advances 4700 BP and the youngest rock glaciers during
one of the two later periods (2000 or 1500 BP). The time
of initiation for the older rock glaciers in FremriGrjótárdalur is more uncertain, but they are most likely
older than the two latest glacial advances and therefore
probably conform to the 3200–3000 BP advance. The
proposed initiation of the landforms also agrees quite
well with the four latest glacier advances in Scandinavia
7.5, 5.1–4.5, 3.2–2.8, 2.2–1.9 and 1.5–1.1 ka BP presented by Karlén (1988) from lacustrine sediments.
Nesje et al. (1991) find evidence for advances at similar
times in Norway, but with a discrepancy attributed to
dating problems and/or differences in topography and
topographic effects.
Based on this discussion we suggest that most of the
landforms in the high mountains of northern Iceland
have developed during the late-Holocene cooling period
after the Atlanticum. The initiation of the oldest forms
(5000 years) coincides with findings from Scandinavia
(Nesje, 1992). The surface ages of around 3000 years
agree with glacier advances found on Svalbard (Svendsen and Mangerud, 1997) and western Greenland
(Kelly, 1985). The surface ages of the different
landforms also coincide with local glacier advance
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
periods described in literature (Stötter et al., 1999). The
surface ages of all investigated landforms in Hóladalur
agree with the Scandinavian glacier and climate
fluctuations found by Karlén (1988).
For Almenningsnöf the streamlines and the discussion of accuracies imply a travel time of around 8000
years, under the assumption of constant movement. But
since the origin of the landform is uncertain, other creep
and landslide processes may have contributed to its
formation, and it is therefore difficult to use the travel
time for surface-age estimates.
6.4. Implications for mountain permafrost distribution
In our discussion the question of permafrost
processes in this maritime, high mountain environment
becomes important. In Hóladalur, the flow analysis
indicates a mixture of glacial flow and additional creep
of the debris.
The mountain areas of Northern Iceland show high
periglacial activity. Whalley and Martin (1994) state that
the Tröllaskagi peninsula has been marginal for
permafrost for the last 200 years and that permafrost
only may appear on the snow-blown summit plateaus at
ca. 1250–1300 m a.s.l. Recent empirical spatial modelling indicates probable mountain permafrost above 850
to 950 m a.s.l. (Etzelmüller, unpublished data). This
means that the landforms in Hóladalur and FremriGrjótárdalur are at or close to the lower level of
permafrost in this area. Considering that the thermal
offset effect of coarse, blocky material, decreases
ground temperature significantly (Harris and Pedersen,
1998), coarse-grained slope bodies will promote the
generation of permafrost. The initial hypothesis of
permafrost conditions is also strengthened by the DTM
analysis and visual inspection of the air photos from
different years of the rock glacier and glacier bodies,
showing that the landform geometry is stabile through
the period from 1964 to 1994. This means that the
ability of the debris layer to protect the ice beneath from
melting during this period has been high for most of the
landforms. Landforms nr. 2 and 3 in Fig. 2, on the other
hand, show signs of partial melt. This could indicate that
the landforms are situated at a level marginal for debris
layer protection of glacier ice in the present climate.
None of the stable landforms is found below 900 m a.s.l.
and the two partially and completely melted landforms
(nr. 2 and 3) are situated between 900 and 1000m a.s.l.,
and 870 and 920 m a.s.l. respectively. It is important to
note that an almost protecting debris layer and ice-cored
moraines that have been stable for some decades do not
necessarily imply permafrost conditions. Krüger and
19
Kjær (2000) report that in the current humid and
subpolar climate it takes 50 years to melt down 40 m of
stagnant dirty ice at Kötlujökull (220 m a.s.l.), an outlet
from Mýrdalsjökull in southern Iceland. The mean
annual air temperature at Kötlujökull is about 2.6 °C
(Krüger and Kjær, 2000) while probably closer to
− 3.0 °C at the front of the studied landforms in
Hóladalur and Fremri-Grjótárdalur at 900 m a.s.l. This
indicates that melt would be even slower here. Everest
and Bradwell (2003) report buried ice that has survived
at least 50 years in front of Virkisjökull and Hrútárjökull, and probably more than 200 years at Skeidaràjökull. All three glaciers are outlets of Vatnajökull in
southern Iceland and even though no temperature
records are given they probably endure a somewhat
milder climate than that found in Hóladalur.
7. Conclusion
The displacement results from Hóladalur suggest that
there is additional creep taking place in the surface of the
debris-covered glacier and the glacier-derived rock
glaciers. Based solely on the displacements it is not
possible to conclude on the cause of this additional
creep. The surface ages of 5000, 3000 and 1500 years
also conform to dating from moraines in the vicinity and
general knowledge of Holocene climate development.
The displacement results from the moving debris
accumulation at Almenningsnöf close to Siglufjörður
agree well with existing GPS surveyed displacements
and add more knowledge on the dynamics of the
landform. Due to the uncertain origin of this landform it
is difficult to relate the results of the travel times of
streamlines to surface age. At Seyðisfjörður the
displacement rates during the investigated 30-year
period were too small to give significant results taking
in account the accuracy of the method, but their uniform
direction implies that movement takes place. The results
of this work show that cross-correlation of orthophotos
is a valuable tool for determining the surface displacement field of selected creeping slope landforms and that
the data can be used for estimating surface ages for some
of these landforms.
Acknowledgements
The project was financially supported by the
Norwegian Research Foundation (NFR), project no.
157837/V30, and the Department of Geosciences,
University of Oslo. Ole Humlum and Leif Sørbel
provided some valuable discussions on the topics
covered in this work. Ole Humlum also gave valuable
ARTICLE IN PRESS
20
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
comments on the manuscript. We also want to thank Jan
Boelhouwers and Ivar Berthling for reviewing the
manuscript.
References
Baltsavias, E.P., 1996. Digital ortho-images—a powerful tool for the
extraction of spatial- and geo-information. ISPRS Journal of
Photogrammetry and Remote Sensing 51, 63–77.
Barsch, D., 1992. Permafrost creep and rockglaciers. Permafrost and
Periglacial Processes 3, 175–188.
Barsch, D., 1996. Rockglaciers. Springer, Berlin. 331 pp.
Berthling, I., Etzelmüller, B., Eiken, T., Sollid, J.L., 1998. The rock
glaciers on Prins Karls Forland, Svalbard (I). Internal structure,
flow velocity and morphometry. Permafrost and Periglacial
Processes 9, 135–145.
Beylich, A.A., 2000. Geomorphology, sediment budget, and relief
development in Austdalur, Austfirðir, East Iceland. Arctic,
Antarctic, and Alpine Research 32, 466–477.
Bindschadler, R.A., Scambos, T.A., 1991. Satellite-image-derived
velocity field of an Antarctic ice stream. Science 252 (5003),
357–364.
Caine, N., 1974. The geomorphic processes of the alpine environment.
In: Ives, J.D., Berry, R.G. (Eds.), Arctic and Alpine Environments.
Methuen, London, pp. 721–748.
Cruden, D.M., Varnes, D.J., 1996. Landslide type and processes. In:
Turner, A.K., Schuster, R.L. (Eds.), Landslides: Investigation and
Mitigation. Special Report, vol. 247. National Research Council
(U.S.). Transportation Research Board 675.
Dahl-Jensen, D., Mosegaard, K., Gundestrup, G.D., Clow, G.D.,
Johnsen, S.J., Hansen, A.W., Balling, N., 1998. Past temperature
directly from the Greenland Ice Sheet. Science 282, 268–271.
Delacourt, C., Allemand, P., Casson, B., Vadon, H., 2004. Velocity
field of “La Clapière” landslide measured by the correlation of
aerial and QuickBird satellite images. Geophysical Research
Letters 31, 15.
Dikau, R., Brunsden, D., Schrott, L., Ibsen, M.-L. (Eds.), 1996.
Landslide Recognition. Identification, Movement and Causes.
Wiley, Chichester. 251 pp.
Everest, J., Bradwell, T., 2003. Buried glacier ice in southern Iceland
and its wider significance. Geomorphology 52, 347–358.
Ferretti, A., Prati, C., Rocca, F., 2000. Non-linear subsidence rate
estimation using permanent scatterers in differential SAR
interferometry. IEEE Transactions on Geoscience and Remote
Sensing 38, 2202–2212.
Ferretti, A., Prati, C., Rocca, F., 2001. Permanent scatters in SAR
interferometry. IEEE Transactions on Geoscience and Remote
Sensing 39, 8–20.
Goldstein, R.M., Engelhardt, H., Kamb, B., Frolich, R.M., 1993.
Satellite radar interferometry for monitoring ice sheet motion;
application to an Antarctic ice stream. Science 262 (5139),
1525–1530.
Guðmundsson, Á., 2000. Frerafjöll og Urðabingir á Tröllaskaga (in
Icelandic). M.Sc. Thesis, University of Iceland. 322 pp.
Guðmundsson, Á., Knudsen, Ó., Hallsteinsson, H., 2003. Seyðisfjörður—Athuganir á setmyndunum og jarðgerð í Þófum og
Botnum. Vetturinn 2002–2003 (report in Icelandic), Jarðfræðistofán—JFS Geological Services, Reykjavik.
Gylfadóttir, S.S., 2003. Spatial interpolation of Icelandic monthly
mean temperature data. 03006, Veðurstofa Islands (The Icelandic
Meteorological Office), Reykjavík.
Haeberli, W., 2000. Modern research perspectives relating to
permafrost creep and rock glaciers. Permafrost and Periglacial
Processes 11, 290–293.
Hamilton, S.J., Whalley, W.B., 1995. Preliminary results from the
lichenometric study of the Nautardálur rock glacier, Tröllaskagi,
northern Iceland. Geomorphology 12, 123–132.
Harris, S.A., Pedersen, D.E., 1998. Thermal regimes beneath coarse
blocky material. Permafrost and Periglacial Processes 9,
107–120.
Humlum, O., 1982. Rock glacier types on Disco Island, Central West
Greenland. Geografisk Tidsskrift 82, 59–66.
Kääb, A., 2002. Monitoring high-mountain terrain deformation from
repeated air- and spaceborne optical data: examples using digital
aerial imagery and ASTER data. ISPRS Journal of Photogrammetry and Remote Sensing 57, 39–52.
Kääb, A., Vollmer, M., 2000. Surface geometry, thickness changes and
flow fields on creeping mountain permafrost: automatic extraction
by digital image analysis. Permafrost and Periglacial Processes 11,
315–326.
Kääb, A., Gudmundsson, G.H., Hoelzle, M., 1998. Surface deformation of creeping mountain permafrost; photogrammetric investigations on Murtel rock glacier, Swiss Alps. In: Lewkowicz, A.G.,
Allard, M. (Eds.), Seventh International Conference on Permafrost,
Proceedings. Collection Nordicana. Centre d'Etudes Nordiques,
Universite Laval, Quebec, PQ, Canada, pp. 531–537.
Karlén, W., 1988. Scandinavian glacial and climatic fluctuations
during the Holocene. Quaternary Science Reviews 7, 199–209.
Kasser, M., Egels, Y., 2002. Digital Photogrammetry. Taylor &
Francis, London. 351 pp.
Kaufmann, V., Ladstädter, R., 2003. Quantitative analysis of rock
glacier creep by means of digital photogrammetry using multitemporal aerial photographs: two case studies in the Austrian Alps.
In: Phillips, M., Springman, S.M., Arenson, L.U. (Eds.),
Permafrost: Proceedings of the Eighth International Conference
on Permafrost. Balkema, Zurich, Switzerland, pp. 525–530.
Kelly, M., 1985. A review of the Quaternary Geology of western
Greenland. In: Andrews, J.T. (Ed.), Quaternary Environments:
Eastern Canadian Arctic, Baffin Bay and Western Greenland.
Allen & Unwin, Boston, pp. 461–501.
Krüger, J., Kjær, K.H., 2000. De-icing progression of ice cored
moraines in a humid, subpolar climate, Kötlusjökull, Iceland.
Holocene 10, 737–747.
Krzystek, P., 1991. Fully Automatic Measurement of Digital Elevation
Models with Match-T, Proceedings of the 43rd Photogrammetric
Week. Schriftenreihe der Universität Stuttgart, pp. 203–213.
Martin, H.E., Whalley, W.B., 1987. A glacier ice-cored rock glacier in
Tröllaskagi, Northern Iceland. Jökull 37, 49–55.
Nesje, A., 1992. Younger Dryas and Holocene glacier fluctuations and
equilibrium-line altitude variations in the Jostedalsbre region,
western Norway. Climate Dynamics 6, 221–227.
Nesje, A., Kvamme, M., Rye, N., Løvlie, R., 1991. Holocene glacial
and climate history of the Jostedalsbreen region, western Norway;
evidence from lake sediments and terrestrial deposits. Quaternary
Science Reviews 10, 87–114.
Østrem, G., 1959. Ice melting under a thin layer of moraine and the
existence of ice cores in moraine ridges. Geografiska Annaler 41A,
228–230.
Þorarinsson, S., Einarsson, T., Kjartansson, G., 1959. On the geology
and geomorphology of Iceland. Geografiska Annaler 41 A,
135–169.
Þordarson, Þ., Hoskuldsson, A., 2002. Iceland. Classic Geology of
Europe, vol. 3. Terra, Harpenden. 200 pp.
ARTICLE IN PRESS
B. Wangensteen et al. / Geomorphology xx (2006) xxx–xxx
Roer, I., Kääb, A., Dikau, R., 2005. Rockglacier kinematics derived
from small-scale aerial photography and digital airborne pushbroom imagery. Zeitschrift für Geomorphologie 49, 73–87.
Rolstad, C., 1995. Satellitt- og flybilder til bestemmelse av
bredynamikk (in Norwegian). M.Sc.-thesis, University of Oslo,
Norway. 129 pp.
Sæmundsson, Þ., 2004. Kortlagning á sigi á Siglufjarðavegi um
Almenninga (report in Icelandic), Vegagerðin, Reykjavik.
Stötter, J., Wastl, M., Caseldine, C., Häberle, T., 1999. Holocene
paleoclimatic reconstruction in northern Iceland: approaches and
results. Quaternary Science Reviews 18, 457–474.
21
Svendsen, J.I., Mangerud, J., 1997. Holocene glacial and climatic
variations on Spitsbergen, Svalbard. Holocene 7, 45–57.
Weydahl, D.J., 2001. Analysis of ERS tandem SAR coherence from
glaciers, valleys, and Fjord Ice on Svalbard. IEEE Transactions on
Geoscience and Remote Sensing 39, 2029–2039.
Whalley, W.B., Martin, H.E., 1994. Rock Glaciers in Tröllaskagi: their
origin and climatic significance. In: Stötter, J., Wilhelm, F. (Eds.),
Environmental Change in Iceland. Münchener Geographische
Abhandlungen, Reihe B, Band, pp. B12289–B12308.
Measuring coastal cliff retreat in the Kongsfjorden area, Svalbard
using terrestrial photogrammetry
B. Wangensteen 1, T. Eiken 1, R. S. Ødegård 2 and J. L. Sollid 1
1. Department of Geosciences, University of Oslo, Norway P.O. Box 1047
Blindern, N-0316 OSLO, Norway.
2. Gjøvik University College, Norway. P.O.Box 191, N-2802 Gjøvik, Norway.
ABSTRACT
Results from two sites for measuring coastal cliff retreat in the Kongsfjorden area on
Svalbard (79˚ N, 13˚ E) are presented. The two sites were established in August 2002
and revisited in August 2004 and show a yearly retreat of 2.7 and 3.1 mm a-1, and these
retreat rates are taken to be significant since most of the retreat takes place within small
areas with rates well above the accuracy. Photos with stereo coverage were taken at
distances of 7 and 15 metres from the cliff walls with a 60 mm Hasselblad camera
mounted on a theodolite. Fixed points were established by drilling bolts into the cliff
wall and surveyed. These fixed points are used as control points for orientation of the
photogrammetric models. The photos were scanned and by the use of digital
photogrammetry a detailed digital terrain model (DTM) is created for each site and year.
The coastal cliff retreat rates are found by differencing the DTMs of 2002 and 2004.
Due to the short distance between camera and cliff, the accuracy of the DTM
differencing is at least 1 cm. The results are analyzed and discussed in light of earlier
rock wall retreat studies in the same area.
KEY WORDS: Digital terrestrial photogrammetry, coastal cliff retreat rates, arctic
coastal dynamics, Svalbard
1. Introduction
The shorelines of the Arctic Ocean compromise about 44,000 km in length
according to the World Vector Shoreline (U.S. Defense Mapping Agency). Over this
distance the coast displays a variety of different coastal types. These coasts also behave
differently when it comes to erosion and sedimentation. In some areas, like the 344 km
of Alaskan coast in the Colville area, the coastal erosion supplies seven times more
sediment to the sea than rivers (Reimnitz et al., 1988). In the Laptev Sea Rachold et al.
(2000) found input from coastal erosion to be twice as big as the river input, while in the
1
Canadian Beaufort Sea the Mackenzie River is the dominant sediment source making
coastal erosion less important (MacDonald et al., 1998). One of the goals of the
international project Arctic Coastal Dynamics (ACD) is to assess the rates and
magnitudes of erosion and accumulation of Arctic coasts and to estimate the amount of
sediments derived from coastal erosion. This work is a part of the ACD project, and the
aim is to get reference data from a relative stable bedrock coastal cliff in the Arctic. The
investigation sites are in the Kongsfjorden area on the western part of the arctic
archipelago Svalbard (Fig. 1).
Fig. 1. Map of Svalbard and the Kongsfjorden
area. The square on the index map represents
the research area, while L is Longyearbyen, T
is Tempelfjorden, W is Wijdefjorden and F is
Prins Karls Forland.
There are very little data available on
costal cliff retreat rates on Svalbard. Prick
(2004) found a rockwall retreat of 1.9 mm a-1
in
a
sandstone
and shale
rockwall
in
Longyearbyen using sediment traps. This used
to be a coastal cliff but it was disconnected from the sea by road and harbour
constructions some 10 years before the monitoring started. Other studies have assessed
general rockwall retreat rates on Svalbard resulting in rates of 0 to 1.58 mm a-1 (Rapp,
1960; André, 1997; Berthling, 2001). All of these studies are based on volumetric
calculation of slope deposit connected with time estimates for the development. In
addition André (1997) directly measured rock wall retreat in areas where both stable
and retreating rockwalls are found using lichonometry as a chronological control.
The ice-foot and sea ice is protecting the coastal cliffs in the fjords on Svalbard
from wave energy during most of the year. Submerged strandflat areas along most of
the outer western coast also reduce the amount of energy reaching the shores (Ødegård
and Sollid, 1993). Arctic coastal cliffs are subjected to different types of subaerial and
marine weathering processes and frost weathering was already proposed by Høgbom
(1914) and Nansen (1922) to be an important process regarding the formation of the
coastal cliffs characteristic of Svalbard fjords. Nansen also suggested that the strandflat
2
forms because of the interaction of frost weathering and marine processes at the base of
the cliffs, with frost weathering as the main fracturing agent and marine processes
removing the weathered material. The strandflat belong to high latitude coasts, and the
genesis has been discusses by e.g. Reusch (1894), Ahlman (1919) and Nansen (1922).
The appearance of the coastal rock walls together with the amount of angular rock
fragments accumulating on the snow and ice-foot below coastal cliffs during spring
show that subaerial mechanical weathering is active and important (O. Liestøl pers.
com) . In some areas the snow and ice could survive all the way through summer above
the high tide, indicating that the removal of material is restricted to heavy storms
recurring in the fjords at several years interval (Ødegård and Sollid, 1993). It has been
observed that the ice-foot plays an important role in both removal of weathered material
from the base of arctic coastal cliffs and fracturing the lower part of cliff walls (Nielsen,
1979).
Since weathered material is removed from the base of the cliff by wave and seaice action, it is impossible to use the weathered deposits for volumetric calculations of
retreat rates. Lichen is very sparse on the faces of the coastal cliffs, denying this as an
option as well. The length of the abrasion platform at the sites is 10-15 metres, but due
to limited knowledge of sea level after the Tapes transgression (Forman et al., 1987) and
with possibilities of the present sea level occupying a pre-Holocene cliff, makes it
difficult to calculate an average Holocene retreat rate (Ødegård and Sollid, 1993).
Satellite remote sensing and aerial photogrammetry have been used for coastal change
studies with success in the Arctic (Grosse et al., 2005; Lantuit and Pollard, 2005;
Mercier and Laffly, 2005). In general digital photogrammetry has become a valuable
tool in geomorphology, using terrestrial or aerial photos depending on the scale of the
feature investigated (Pyle et al., 1997; Chandler, 1999; Lane et al., 2000; Baily et al.,
2003; Chandler et al., 2003; Lim et al., 2005). Because retreat rates of coastal cliffs in
Kongsfjorden are expected to be small, satellite remote sensing and aerial
photogrammetry are not able to provide the required precision for the time scale
available. However, terrestrial photogrammetry with camera/object distances around 10
m was chosen as the measuring technique since it will provide the millimetre to
centimetre accuracy needed to measure the expected coastal cliff retreat rates. The aim
of this study is therefore to quantify the retreat rates of two coastal cliff sites in the
3
Kongsfjorden
area
using
multi-temporal
terrestrial
photography
and
digital
photogrammetry.
2. Study area
The Kongsfjorden area is situated on the western coast of Spitsbergen, the
largest island of the Svalbard archipelago (Fig. 1). Svalbard is situated between 74q 81q N and 10q - 35q E in the North Atlantic, and has permafrost conditions down to sea
level (Liestøl, 1977). Glaciers of different types cover 60% of the 66,000 km2
archipelago (Hagen et al., 1993). The mean annual air temperature (MAAT) in NyÅlesund is -6˚C and the mean annual precipitation 370 mm (Førland and HanssenBauer, 2000). The tidal range in Ny-Ålesund is 1.8 m having a maximum tidal range of
2.3 m. The average high tide is 0.6 m above mean sea level (Ødegård and Sollid, 1993).
The area has continuous permafrost with measured depths ranging from 130 to 150 m
near the shores of Ny-Ålesund (Orvin, 1944; Liestøl, 1977) . The mean annual ground
temperature (MAGT) is believed to be about -5˚C based on measurements in a former
coal mine (Orvin, 1944).
The two sites investigated in this study (Fig. 1) are both close to Ny-Ålesund
and consist of low cliffs in bedrock with unconsolidated material on top, and with sandy
and stony beaches in front (Ødegård et al. 1987). Site 2 is located within a small bay
and is less exposed than site 1 located further out the fjord with a much longer fetch.
The bedrock of both investigated sites is made up of dolomitic limestone of Middle
Carboniferous to Early Permian age. The unconsolidated material of 1 to 2 m in
thickness found on the top of the bedrock cliffs are marine deposits of Quaternary age
(Hjelle et al., 1999). The dimension of the area analysed at site 1 is 2.8 by 3.6 metres,
and at site 2 it is 2.2 by 2.4 m. The investigations are limited to the bedrock part of the
cliffs.
Coastal geomorphology maps of Svalbard, excluding Nordaustlandet, reveal
well- developed coastal cliffs along 27% of the coastline and another 13% with rocky
shores (Etzelmüller et al., 2003). Low vertical cliffs are most frequent in the fjord areas
in northwestern Spitsbergen. This means that the chosen sites it that manner are
representative for a great portion of Svalbard’s coastline and an even greater portion of
the fjord areas. However, one should bear in mind that bedrock properties and marine
4
processes vary greatly within the same coastal type. The percentage of artic coasts that
are cliffed and rocky is not known, but these coast types compromise three-quarters of
the world’s coastline (Bird, 2000).
Ødegård and Sollid (1993) investigated the temperature in a coastal cliff near
Ny-Ålesund on an hourly basis from August 1987 to August 1988. The steep
temperature gradients and available water from snow melting are believed to favour
water migration in the fractured dolomitic limestone at below zero temperatures. Spring
and summer snowmelt periods are therefore found to be favourable for crack expansion
and rock fracturing. A similar study was performed in Liefdefjorden from May to
August 1992, and a corresponding similar temperature regime was found (Ødegård et
al., 1995). It is suggested that permafrost controls the cliff temperature regime in such a
way that the conditions in spring and summer with a melting snow cover providing
sufficient moisture, is favourable to rock break down. Ødegård and Sollid (1993) and
Ødegård et al.(1995) did not perform any direct measurement of crack expansion or
retreat rates in their studies.
When it comes to rockwall retreat rates, André (1997) reports of the greatest
retreat of up to 1.58 mm a-1 for the area in the end of Kongsfjorden and addresses only
0.11-0.22 mm a-1 of this to frost shattering and the rest to post-glacial stress relaxation.
This leaves cryogenic weathering as the most important process acting in the current
environment.
3. Photo acquisition and DTM differencing
Photographs with stereo coverage were acquired with a 60 mm analog
Hasselblad camera at each site in August 2002 and August 2004. A camera/object
distance of 15 and 7 m was used at sites 1 and 2 respectively. The geometry of the
camera set up is shown in Fig. 2. At both sites drilling a bolt into bedrock outcrop and
measuring its position by relative GPS established a fixed point for global reference.
The positions of the camera stations were also measured using relative GPS and by
theodolite and electronic distance measurer from the fixed point to tie both a global and
a local reference. At both sites bolts to be used as photogrammetric control points were
drilled into the cliff wall as well. These fixed points were surveyed from the camera
5
Fig. 2. The camera
setup used.
stations and the camera was mounted on top of the theodolite with a special device. This
makes it possible to ensure parallel photo directions.
The photos where scanned at a resolution of 6.25 Pm (4,000 dpi) and imported
into the Z/I Imagestation digital photogrammetric workstation (DPW) software. A local
coordinate system was established to simulate aerial photogrammetry, with the z-axis is
pointing out of the cliff and the xy-plane vertical behind the cliff wall and also parallel
to the photo base (Fig. 2). All coordinates were also scaled by a magnitude of 100 to get
reasonable “flying heights”. The photo pairs have been used to create photogrammetric
stereo models using the surveyed pass points. A high precision digital terrain model
(DTM) was generated for each site using the automatic elevation point collector module
in the DPW. To ensure high quality of the DTMs, the photogrammetric models from
different years at the same site were tied together using the same control points and five
additional tie points located in stable parts of the cliff. The retreat rate was simply found
by differencing the 2002 and 2004 models using ESRI’s geographic information system
ArcInfo Workstation.
Using terrestrial digital photogrammetry with a similar approach and evaluating
the result Pyle et al. (1997) infer a DTM accuracy of 12 mm with 15 m camera/object
distance and point out that this accuracy of 1/1,000 of the camera/object distance is in
the lower end of the accuracies achieved by aerial photogrammetry. This relatively large
relative error is attributed to errors in the estimates of ground point positions being
much larger relative to the camera/object distance for close range photogrammetry.
Using a digital camera and a 1.6 m camera/object distance Chandler et al (2003) found
6
an elevation RMS error of 1.9 mm, corresponding to 12 parts per 10,000. No direct
evaluation of the DTM accuracy was performed in this study, but it is reasonable to
believe that they are in the proximity of the findings of Pyle et al. (1997). For
differentiation of two DTMs, the accuracy found by Pyle et al. (1997) for each
individual model would lead to a total error for the differentiation of 10.6 mm a-1 for
site 1 and 4.9 mm a-1 at site 2.
The RMS values calculated during the combined absolute orientation of the
2002 and 2004 models within the DPW is 3 mm, 4 mm and 3 mm for X, Y and Z
respectively at site 1 and 2 mm, 3mm and 6 mm at site 2. This means that the accuracy
should be 10 mm or better for the yearly retreat values. Photos were also acquired at
two additional sites without using fixed points for exterior orientation. Due to lack of
success in the orientation procedure results from these sites are not presented.
Fig 3. DTMs of a) site 1 and b)
site 2 from 2004 as shadowed
relief models.
7
4. Coastal cliff retreat rates
The automatically generated 2004 DTMs for site1 and site 2 are shown in Fig. 3
as shadowed relief models. The differencing of the DTMs reveal an average retreat of
3.1 mm a-1 for site 1 and 2.7 mm a-1 for site 2 (Fig. 4). Both difference models show a
spot like retreat with maximum retreat rates of 12.5 cm a-1 and 4.2 cm a-1 for site 1 and
site 2 respectively. It can be seen that several of the areas of greatest retreat at site 1 are
found around overhanging areas. The other retreat spots at both site 1 and site 2 are in
areas where the cliff is quite planar, and therefore show flaking rather than small rock
falls (Fig. 5). When comparing orthophotos of site 1 from 2002 and 2004, they show
removal of material in front of the cliffs between the two acquisitions. The volume of
the removed boulders along the
6.5
m
distance
with
stereo
coverage is estimated to be about
0.6 m3 based on photogrammetric
stereo measurements. Since the
boulders are placed in front the
cliff it is not possible to measure
them using DTM differencing and
this area is therefore excluded
from
the
DTM
differencing
presented in Fig. 4. The blocky
material that has been removed
from the beach at site 1 is also
marked in Fig. 5.
Fig 4. Retreat rates from a) site 1
and b) site 2 draped over the 2004
shadowed relief models.
8
Fig. 5 Orthophoto of site 1 in 2004
showing areas of retreat and block
removal.
5. Discussion
5.1 DTM differencing accuracy
Looking at RMS values from the orientation of the model or using theoretical
assumptions on the relation between camera/object distance and accuracy, the accuracy
of the applied method is 10 mm or better for yearly retreat rates at both sites. As an
additional check of the accuracy, semivariograms were made for the retreat rate grids.
In these grids all points having retreat rates above 1 cm were left out. Semivarograms
are used in geostatistics to evaluate spatial dependency in data sets (Burrough, 1990)
and are plots of the semivariance as a function of distance. The difference between the
nugget (i.e a measure for variance caused by noise in the data) and the sill (i.e. a
measure for the variance level at the distance where the spatial dependence cease) can
be taken as an indicator for the overall spatial dependency in the data. At both sites
there was a relatively small difference between the sill and the nugget value, namely 4%
for site 1 and 20% for site 2, meaning that there are only small spatial dependency in the
dataset for values below the believed accuracy and that the retreat rates should be valid
as volume measures and not only point measures . No spatial trends in the datasets
where found either. It is also worth noting that the semivariograms and the RMS values
from the orientation show that the dataset from site 1 is more accurate than at site 2,
even though site 1 has twice the camera/object distance compared to site 2. This might
9
be due to fewer fixed points being used for orientation at site 2, and hence a less
accurate orientation of the models.
As noted earlier most of the retreat detected in this study takes place in
concentrated and rather small areas and with a retreat rate greater than the accuracy,
meaning that the overall retreat rates also should be significant.
5.2 Retreat rates
Compared to the findings of Rapp (1960), André (1997), Berthling (2001) and
Prick (2004) the results from this study show somewhat higher retreat rates. André
(1997) found retreat rates up to 1.58 mm a-1, but attributes only a small portion of these
rates to currently active frost shattering, namely 0.070 mm a-1 in average for
amphibolite in Wijdefjorden and 0.163 mm a-1 for quartzite in Kongsfjorden. Rapp
(1960) reports of a present rock wall retreat rate of 0.02 to 0.2 mm a-1 for the sand and
limestone in Tempelfjorden and for a nunatak in the same area the higher value of 0.5
mm a-1 is presented. Berthling (2001) reports on rock wall retreat rates of 0.3 to 0.6 mm
a-1 based on volume and age estimates of rockglaciers on Prins Karls Forland in an area
of quartzite bedrock. None of these localities are exposed to marine processes and
therefore not directly comparable with the results from this study. On the other hand,
Prick (2004) reports a rate of 1.9 mm a-1 for the sandstone and shale cliff wall in
Longyearbyen and this cliff is the most similar of earlier investigated localities because
of similarities in process and bedrock composition. The differences between the results
found in this study and the earlier works of Rapp (1960), André (1997) and Berthling
(2001) can be attributed to differences in the effectiveness of the rock wall retreat
processes involved compared to coastal processes. The smaller retreat rate found by
Prick (2004) may be explained by differences in bedrock and that the location used in
her study consists of bedrock that is somewhat less susceptible to weathering. But
according to Sunamura (1992) cliffs consisting of shale and sandstone should be more
susceptible to coastal erosion than limestone. The locations used in this study are
currently influenced by marine processes while the site of Prick (2004) has been
protected from these processes during a decade, which could lead to observed
differences. André (1997) attributes the differences between her findings of non-coastal
rock wall retreat at Svalbard and findings in Alpine regions, showing current retreat
10
rates of 1 to 3 mm a-1, to differences in bedrock and bedrock stresses due to tectonics.
André (1997) also states that similar high rates as those found in Alpine areas were the
case at Svalbard during the post-glacial stress relaxiation.
Marine cliffs in general show a variety of erosion rates caused by lithology, from
several metres per year for regolith, via 4 mm a-1 and upwards for rocks of medium
hardness to hard-rock cliffs unaffected by erosion (Young and Saunders, 1986).
Sunamura (1992) lists worldwide linear cliff recession rates and found average rates for
granite of 1 mm a-1, 1 to10 mm a-1 for limestone and 1 cm a-1 for shale. Allard and
Trembelay (1983) found erosion rates of a basaltic bed rock coast in the order of 1 cm a1
by Hudson Bay in northern Quebec, Canada and attributes this to the joint actions of
waves, sea-ice and frost shattering. Since carbon dioxide is more soluble at colder
temperatures one should expect higher rates of chemical erosion in arctic marine
environments (Williams, 1949). It is therefore also reasonable to believe that chemical
weathering is more active in the coastal zone compared to areas not influenced by the
sea. At a non-coastal locality chemical erosion rates of 2mm ka-1 has been found while
the rates in marine polar areas may be at least two magnitudes greater (Corbel, 1959).
This means that chemical weathering only would explain a negligible portion of the
coastal cliff retreat rates measured in this study.
It is important to bear in mind that the measurements that form the base of this
study is done only at two selected localities and compromise only some metres of
coastline. Since the processes governing coastal cliff retreat probably vary considerably
both spatially and temporally, due to variations in sea-ice cover, snow cover, bedrock,
existence of the ice-foot together with wave activity and storminess, the retreat rates
will also show a huge variation. Earlier studies for instance show a considerable spatial
and temporal variation in sea-ice and ice-foot conditions in the Kongsfjorden area
(Wiseman et al., 1981) and other places on Svalbard (Feyling-Hansen, 1953). To assess
reliable retreat rates a long-term programme, preferably also including more sites,
should be undertaken. Monitoring of snow and sea ice cover at the investigated sites
would also be an advantage.
Traditional geomorphologic estimates of coastal retreat rates in the scale of
millimetre or centimetre per year are based on a longer period of time than two years,
and the uncertainty of those are mainly due to uncertainties of the time horizon (i.e.
11
dating). The advantage of using volume/time based retreat rates is that the year-to-year
variation in the processes governing weathering and coastal processes is better dealt
with. As noted earlier the uncertainty of the sea level development since the Tapes
transgression in the investigated area makes it difficult to estimate the Holocene retreat
rate using these traditional methods (Forman et al., 1987). However, terrestrial
photogrammetry makes is possible to perform spatially detailed studies of retreat rates
and to map more short-term temporal variations.
6. Conclusion
The method of close up digital terrestrial photogrammetry seems suitable for
creating accurate DTMs of coastal cliffs and for calculating retreat rates in the order of
millimetres to centimetres. Differencing of multi-temporal DTMs show reveal retreat
rates of 3.1 mm a-1 and 2.7 mm a-1 for two costal cliff locations in the Kongsfjorden
area on western Spitsbergen, Svalbard. The retreat has a spot like appearance and the
processes involved seem to be both flaking and small rock falls. Compared to rock wall
retreat rates found elsewhere on Svalbard the rates are quite high. This could be due to
other investigations being performed in non-coastal environments or the natural
variation in the processes involved. Two years is too short a time interval of
investigating coastal cliff retreat rates. Nevertheless, the retreat rates found are within
the range of worldwide averages for limestone cliffs. Removal of weathered material at
the base of the coastal cliffs is considered to be important in the retreat process and is
documented through the multi-temporal photography obtained at one of the sites. About
0.6 m3 of material is removed from the 6.5 m wide area covered by the stereo model at
one of the investigated sites.
Acknowledgements
The fieldwork was financially supported by the INTAS-project Arctic coasts of
Eurasia: dynamics, sediment budget and carbon flux in connection with permafrost
degradation (INTAS-2001-2329) and the Norwegian Research Council on behalf of the
Norwegian Polar Committee .
12
References
Ahlman, HW. 1919. Geomorphical studies in Norway. Geografisk Annaler 1: 3-252.
Allard, M and Tremblay, G. 1983. Les processus d'érosion littorale périglaciaire de la
région de Poste-de-la-Baleine ed des îles Mnaitounuk sur la côte est de la mer
d'Hudson, Canada. Zeithschrift für Geomorphologie, Supplementband 47: 27-60.
André, M-F. 1997. Holocene Rockwall Retreat in Svalbard: A triple-rate evolution.
Earth Surface and Processes and Landforms 22: 423-440.
Baily, B, Collier, P, Farres, P, Inkpen, R and Pearson, A. 2003. Comparative assessment
of analytical and digital photogrammetric methods in the construction of DEMs
of geomorphological forms. Earth Surface and Processes and Landforms 28:
307-320.
Berthling, I. 2001. Slow periglacial mass wasting - processes and geomorphological
impact. Case studies form Finse, southern Norway and Prins Karls Forland,
Svalbard. Series of dissertations submitted to the Faculty of Mathematics and
Natural Sciences. University of Oslo. Oslo. Dissertation for the Degree of Dr.
Scient
Bird, E. 2000. Coastal geomorphology: an Introduction. Wiley, Chichester
Burrough, PA. 1990. Principles of Geographical Information Systems for Land
Resource Assessment. Clarendon Press, Oxford
Chandler, J. 1999. Effective application of automated digital photogrammetry for
geomorphological research. Earth Surface and Processes and Landforms 24: 5163.
Chandler, JH, Buffin-Bélanger, T, Rice, S, Reid, I and Graham, DJ. 2003. The accuracy
of river a bed moulding/casting system and the effectiveness of a low-cost
digital camera for recording river bed fabric. Photogrammetric Record 18: 209223.
Corbel, J. 1959. Vitesse de l'érosion. Zeithschrift für Geomorphologie 3: 1-28.
Etzelmüller, B, Ødegård, RS and Sollid, JL. 2003.The spatial distribution of coast types
on Svalbard. In Arctic Coastal Dynamics Report of the 3rd international
workshop University of Oslo (Norway) 2-5 December 2002, Rachold, V, Brown,
J, Solomon, S & Sollid, JL (eds). Alfred Wegener Institute for Polar and Marine
Research: Bremerhaven.
Feyling-Hansen, RW. 1953. Brief account of the ice-foot. Norsk Geografisk Tidsskrift Norwegian Journal of Geography 14: 45-52.
Forman, S, Mann, DH and Miller, GH. 1987. Late Wechselian and Holocene sea-level
history of Brøggerhalvøya. Spitsbergen. Quaternary Research 27: 41-50.
Førland, EJ and Hanssen-Bauer, I. 2000. Increased precipitation in the Norwegian
Arctic; true or false?. Climatic change 46: 485-509.
Grosse, G, Schirrmeister, L, Kunitsky, VV and Hubberten, H-W. 2005. The use of
CORONA images in remote sensing of periglacial geomorphology: An
illustration from the NE Siberian coast. Permafrost and Periglacial Processes
16: 163-172.
Hagen, JO, Liestøl, O, Roland, E and Jørgensen, T. 1993. Glacier Atlas of Svalbard and
Jan Mayen. Norsk Polarinstitutt Meddelelser. 129. Norsk Polarinstitutt, Oslo
Hjelle, A, Piepjohn, K, Saalmann, K, Ohta, Y, Salvigsen, O, Thiedig, F and Dallmann,
WK. 1999. Geological map of Svalbard 1:100,00 Sheet AG7 Kongsfjorden.
Norwegian Polar Institute, Tromsø.
13
Høgbom, B. 1914.Ûber die geologische bedeutung des frostes. In Bulletin of the
Geological Institution of the University of Upsala, Sjøberg, H (ed.)294-306.
Lane, SN, James, TD and Crowell, MD. 2000. Application of digital photogrammetry to
complex topography for geomorphological research. Photogrammetric Record
16: 793-821.
Lantuit, H and Pollard, WH. 2005. Temporal stereophotogrammetric analysis of
retrogressive thaw slumps on Herschel Island, Yukon Territory. Natural
Hazards and Earth System Sciences 5: 413-423.
Liestøl, O. 1977. Pingos, springs, and permafrost in Spitsbergen. Norsk Polarinstitutt
Årbok 1975 7-29.
Lim, M, Petley, DN, Rosser, NJ, Allison, RJ, Long, AJ and Pybus, D. 2005. Combined
digital photogrammetry and time-of-flight laser monitoring for cliff evolution.
Photogrammetric Record 20: 109-129.
MacDonald, RW, Solomon, S, Cranston, RE, Welch, HE, Yunker, MB and Gobiel, C.
1998. A sediment and organic carbon budget for the Canadian Beaufort Shelf.
Marine Geologogy 144.
Mercier, D and Laffly, D. 2005.Actual paraglacial progradation of the coastal zone in
the Kongsfjorden area, western Spitsbergen (Svalbard). In Cryospheric Systems:
Glaciers and Permafrost, Harris, C & Murton, JB (eds). Geological Society:
London; 111-117.
Nansen, F. 1922. The strandflat and isostasy. Videnskapsselskapets skrifter I. MatNaturv. klasse 1921. 2.
Nielsen, N. 1979. Ice-foot processes. Observations of erosion on a rocky coast, Disko,
West Greenland. Zeithschrift für Geomorphologie 23: 321-331.
Ødegård, RS and Sollid, JL. 1993. Coastal cliff temperatures related to the potential for
cryogenic weathering processes, western Spitsbergen, Svalbard. Polar Research
12: 95-106.
Ødegård, RS, Etzelmüller, B, Vatne, G and Sollid, JL. 1995.Near-surface spring
temperatures in an Arctic coastal cliff: Possible implications of rock breakdown.
In Steepland geomorphology, Slaymaker, O (ed.) Wiley: Chichester; 89-102.
Orvin, AK. 1944. Litt om kilder på Svalbard (in Norwegian). Norsk Geografisk
Tidskrift - Norwegian Journal of Geography 10.
Prick, A. 2004. Observations of rock temperatures and rock moisture variability in
Longyearbyen: Implications for cryogenic weathering and rock wall retreat rate
Abstracts of Pace 21: Permafrost and Climate in the 21st Century. Field
Workshop Longyearbyen, Svalbard, 8-13 September 2004 16.
Pyle, CJ, Richards, KS and Chandler, JH. 1997. Digital photogrammetric monitoring of
river bank erosion. Photogrammetric Record 15: 753-764.
Rachold, V, Grigoriev, MN, Are, FE, Solomon, S, Reimnitz, E, Kassens, H and
Antonow, M. 2000. Coastal erosion vs. riverine sediment discharge in the Arctic
shelf seas. International Journal of Earth Sciences (Geol. Rundsch.) 89: 450460.
Rapp, A. 1960. Talus slopes and mountain walls at Tempelfjorden, Spitsbergen - a
geomorphological study of the denudation of slopes in an arctic locality.
Meddelanden Uppsala Universitets Geografiska Institution. 55A.
Reimnitz, E, Graves, SM and Barnes, PW. 1988. Beaufort Sea coastal erosion, sediment
flux, shoreline evolution and erosional shelf profile. Map I-1182-G and text.
U.S. Geological Survey:
14
Reusch, H. 1894. Strandflaten, et nyt træk i Norges geografi. Norges Geologiske
Undersøgelse 14: 1-14.
Sunamura, T. 1992. Geomorphology of rocky coasts. Coastal morphology and research.
Wiley, Chichester
Williams, JE. 1949. Chemical weathering at low temperatures. Geographical Review
39: 129-135.
Wiseman, WJ, Owens, EH and Kahn, J. 1981. Temporal and spatilal variability of icefoot morphology. Geografisk Annaler 63 A: 69-80.
Young, A and Saunders, I. 1986.Rates of surface processes and denuation. In Hillslope
processees, Abrahams, AD (ed.) Allen & Unwin: Bostoun; 3-27.
15
SURFACE ELEVATION CHANGE AND HIGH RESOLUTION SURFACE VELOCITIES FOR ADVANCING OUTLETS OF
SURFACE ELEVATION CHANGE AND HIGH
RESOLUTION SURFACE VELOCITIES FOR
ADVANCING OUTLETS OF JOSTEDALSBREEN
BY
BJØRN WANGENSTEEN 1, OLE MAGNUS TØNSBERG 1, ANDREAS KÄÄB 2,
TROND EIKEN 1 AND JON OVE HAGEN 1
1
Department of Geosciences, University of Oslo, Norway
of Geography, University of Zürich-Irchel, Zürich, Switzerland
2Department
Wangensteen, B., Tønsberg O.M., Kääb, A., Eiken T. and Hagen,
J.O., 2006: Surface elevation change and high resolution surface
velocities for advancing outlets of Jostedalsbreen. Geogr. Ann.,
88 A (1): 55–74.
ABSTRACT. Velocity fields for the three outlet glaciers Nigardsbreen, Bergsetbreen and Baklibreen of
Jostedalsbreen, southern Norway, have been obtained from successive orthophotos by digital image
processing. The orthophotos were generated with
standard digital photogrammetry. They are based on
air photos acquired ten days apart in August 2001
and produced with a ground resolution of 0.5 m. Glacier displacements were calculated by matching homologous points in each orthophoto using a regular
grid with 10 × 10 m resolution. Displacements, calculated using cross-correlation matching of orthophotos for Nigardsbreen, show daily movement of up to
1.34 m and agree well with GPS-measured velocities
during the same period. The average velocities for
the three glaciers ranged from 0.38 to 0.56 m per day.
Surface elevation change have been calculated by
differencing the 5 m resolution digital elevation
model (DEM) obtained from the August 2001 air photos with two 25 m resolution DEMs from 1984 (Nigardsbreen) and 1993 (all three outlets). These calculations show an average increase in surface elevation of 22.1 m for Nigardsbreen between 1984 and
2001. Bergsetbreen and Baklibreen show increases
of 3.2 m and 14.3 m between 1993 and 2001. In addition a DEM produced from air photos from 1997 was
used for Nigardsbreen, making it possible to show
how the surface elevation has changed between the
years 1984, 1993, 1997 and 2001. The advance of the
glacier snouts is also clearly visualized by the DEM
differencing
Key words: digital photogrammetry, glacier velocities, orthophotos, volume change
Introduction
During the last few decades most glaciers in the
world have been retreating (IAHS(ICSI) et al.
1998, 2001). As a widely recognized exception to
this global trend, the glaciers in the western part of
Norway have been advancing since the late 1980s
until the start of this millennium. This advance was
related to increased winter precipitation in the late
1980s and beginning of the 1990s (Kjøllmoen
2003b, 2004). In order to survey both velocity and
surface elevation changes on the outlets of Jostedalsbreen, air photos covering ten outlets of this
large glacier were acquired on 19 and 29 August
2001. The temporal resolution of ten days was considered feasible for mapping the expected glacial
displacements, based on the precision of the method used.
The first objective of this study was to better understand the glacier-dynamic processes associated
with the above-mentioned glacial advance. In addition, the study aimed to evaluate, for the first
time, the applicability of high-resolution digital
image matching techniques for deriving glacier
surface velocities over short time intervals from repeated aerial photography. Therefore, the flow
fields of two steep outlet glaciers, Bergsetbreen and
Baklibreen, as well as one valley glacier, Nigardsbreen, were mapped using photogrammetric orthophoto generation and cross-correlation matching of
orthophotos from different dates. Digital elevation
models (DEMs) existed from 1984 and 1993, and
two new ones were generated with aerial photographs of 1997 and 2001 for Nigardsbreen. The
1993 and 2001 DEMs are the only ones covering
Bergsetbreen and Baklibreen. An assessment was
made on the volume change of the glaciers by differencing these multi-temporal DEMs using a geographical information system (GIS).
Several of the outlets of Jostedalsbreen and especially Nigardsbreen have been the subject of
many glaciological studies in the past which has
Geografiska
Annaler
· 88 A (2006) · 1
©
The authors
(2006)
Journal compilation © (2006) Swedish Society for Anthropology and Geography
55
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig.1. Location of Jostedalsbreen in southern Norway and location of the study area (rectangle).
Source: © Statens kartverk
led to comprehensive sets of data: length change
dating back to 1748, mass balance dating back to
1962, and velocity measurements from several of
the decades since the 1930s are available (Østrem
et al. 1976). A GPS measurement campaign for
glacier velocity was undertaken in 2001 and 2002
(Tønsberg 2003), covering the period in late August 2001 when the air photos used in this study
were acquired. These data serve as a solid base for
comparison and validation of the results from this
study together with the earlier work mentioned
above.
Nigardsbreen, Baklibreen and Bergsetbreen
All three investigated glaciers, Nigardsbreen
(61°41'N, 7°11'E), Baklibreen (61°40'N, 7°5'E)
and Bergsetbreen (61°39'N, 7°4'E), are outlets of
56
Jostedalsbreen (487 km2), the largest glacier on
mainland Europe (Østrem et al. 1988) situated in
the western part of southern Norway (Fig. 1). All
three outlets used in this study are located on the
southeastern side of the ice cap. Jostedalsbreen is a
maritime glacier exemplified by Nigardsbreen,
with a mean specific winter balance of 2.4 m water
equivalents, and a mean equilibrium line altitude of
1506 m a.s.l. in the period 1962–2003 (Kjøllmoen
2003b, 2004). During winters of high precipitation
the winter balance has reached levels of 5.6 m water
equivalents (1989) in the upper areas of Nigardsbreen (Østrem et al. 1991).
Nigardsbreen is a valley glacier, 9.6 km in
length, which drains a large area of the ice cap and
flows down into a deep U-shaped valley through
three different tributary icefalls (Fig. 2). Baklibreen
is a smaller and much steeper glacier (Fig. 2). The
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 2. Map of the three investigated outlets Nigardsbreen (1), Baklibreen (2) and Bergsetbreen (3).
The coverage of the orthophotos
used is shown
Source: Statens kartverk ©
lower part of Baklibreen is a hanging glacier with
up to 33° inclination situated on the valley side of
Krundalen, just above the snout of Bergsetbreen.
Bergsetbreen drops steeply into the valley end of
Krundalen having a slope angle of 27°, which continues over an area covering 1000 vertical metres
(Fig. 2). The orthophoto coverage used in this study
is shown in Fig. 2. Nigardsbreen extends from 355
to 1950 m a.s.l., Baklibreen from 950 to 1950 m
a.s.l., and Bergsetbreen from 560 to 1960 m a.s.l.
(Østrem et al. 1988).
Earlier investigations
Two of the glaciers have length change data: Nigardsbreen has a continuous record from 1899 and
Bergsetbreen has records between 1899 and 1945,
and from 1996 and onwards. For Nigardsbreen
there are also historical data on glacier length dating back to 1748 (Østrem et al. 1976). In addition,
Nigardsbreen has a mass balance data series, the
longest for the Jostedalsbreen outlets and the second longest in Norway, dating back to 1962.
Nigardsbreen had a continuous positive mass
balance in the majority of the years from 1964 to
2000 and gained 17 m water equivalents from 1962
to 2003; however, it still continued to retreat rapidly until the 1970s. From 1987 to 2003 Nigardsbreen
Geografiska Annaler · 88 A (2006) · 1
advanced 270 m. This advance was explained by
increased winter precipitation in the late 1980s and
the beginning of the 1990s (Andreassen et al.
2006). The overall net retreat for Nigardsbreen
from measurements started in 1899 is 2300 m, and
the net retreat from the Little Ice Age maximum in
1748 until 1899 was calculated at 1995 m (Østrem
et al. 1976). Since 1899 Bergsetbreen has shown a
net retreat of 320 m; however, it is known to have
advanced rapidly just before length change measurements started again in 1996 (Andreassen et al.
2006). Surface elevation and velocity of Baklibreen was monitored from 1987 to 1999 (Kjøllmoen 2000) and surface elevation again from 2001
to 2003 (Kjøllmoen 2004) due to an ice avalanche
killing three hikers in 1986.
Concerning glacier velocities Nigardsbreen is
by far the most investigated of the three outlets.
Velocity measurements for Nigardsbreen were
conducted and recorded as early as 1937 and 1938
by a German expedition (Pillewizer 1950) using
the terrestrial photogrammetric method introduced by Finsterwalder (1931). Liestøl used a
similar method in 1949, 1951, 1953 and 1961
(Østrem et al. 1976); and trigonometric stake
measurements were made from 1966 to 1969
(Nielsen 1970). The previous investigations revealed velocities ranging from a few centimetres
57
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Table 1. Data on the digital elevation models used
Photo date
10 Aug. 1984
8 Sep. 1993
14 Aug. 1997
29 Aug. 2001
Resolution
(m)
Flying height
(m.a.s.l)
25
25
5
5
6300
6150
6150
3950
Height accuracy
(RMS m)
5
5 (4–6)
1.3*
0.8*
Coverage
(glaciers)
Nigardsbreen
All
Nigardsbreen
All
Source†
NVE
SK
UoO
UoO
*Height accuracy for the 1997 and 2001 DEMs is calculated based on the flying height, an average terrain elevation of 850 m a.s.l. with
a height accuracy of c. 0.025% of the flying height (above terrain) using error propagation
†NVE, Norwegian Water Resources and Energy Directorate; SK, Norwegian Mapping Authority; UoO, the Department of Geosciences,
University of Oslo.
per day at the tongue up to 1.4 m per day (md–1)
in one of the icefalls.
Tønsberg (2003) reports on increased velocities
on the tongue of Nigardsbreen in later years (2001–
2002) compared to reports from 1950–1970 and attributes this change to the increased thickness and
width of the tongue. Using differential GPS, he
found velocities ranging from 0.22 to 0.80 m d–1 in
different periods during 2001 and 2002 in the elevation range of 460 to 790 m a.s.l. He also found the
velocity to be increasing with elevation and to be
higher during the summer season, especially in the
lower parts of the tongue.
Methods
Generation of DEM and orthophotos
Cross-correlation matching has been used for mapping glacier velocities in satellite imagery since the
early 1990s (Bindschadler and Scambos 1991) and
in aerial orthophotos since 1995 (Rolstad 1995).
Aerial photogrammetry has been used as an effective tool for velocity and geometry change measurements of glaciers, landslides and creeping permafrost landforms such as rock glaciers for several
decades (Finsterwalder 1931; Haeberli et al. 1979;
Kääb et al. 1997; Kääb and Funk 1999), and since
the advent of digital photogrammetry cross-correlation matching of orthophotos has also been widely used (Baltsavias 1996; Kääb and Vollmer 2000;
Kaufmann and Ladstädter 2003; Delacourt et al.
2004)
In this study orthophotos were generated using
a Z/I-Imaging digital photogrammetric workstation (DPW) with air photos acquired on 19 and 29
August, 2001 by Fotonor. The air photos have a 1:
20 000 scale (flying height of 3100 m above mean
ground level) and are scanned with 14 μm resolution giving 0.3 m geometrical resolution. Ortho58
photos are air photos transformed to orthogonal
projection (i.e. map projection) by the use of a digital elevation model (DEM) and a DPW (Kasser
and Egels 2002). Having a stereo model it is possible to automatically generate a DEM using the
digital image matching module in the DPW. An orthophoto is then generated for each date for each
glacier by resampling of the scanned aerial photographs using the orientation parameters and the
corresponding DEM. The orthophotos were generated with a 0.5 m resolution using DEMs of 5 m resolution.
Cross-correlation of orthophotos
For measurement of local horizontal displacements
of the glaciers, cross-correlation matching of orthophotos was used using the CIAS-software
(Kääb and Vollmer 2000). This software matches
homologous (conjugate) points in two geo- and coreferenced orthophotos of the same area taken at
different times. The points are selected as a regular
grid with 10 m spacing in the orthophoto of time 1.
A small reference window is extracted from the orthophoto of time 1 around each point and the homologous point to the centre point of this small
window is searched for in a larger test area around
the corresponding coordinate in the orthophoto of
time 2. A cross-correlation factor is calculated for
each possible location of the reference window
within this test area. The location that yields the
highest correlation factor is taken to be the position
of the homologous point in the orthophoto of time
2. For a thorough description see Kääb and Vollmer
(2000). If displacement has taken place during the
time interval between the two acquisitions, then the
displacement is measured as the horizontal coordinate distance between the positions of the two homologous points (this method does not measure the
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Table 2. Maximum and average velocities measured by cross-correlation of the 19 and 29 August
2001 orthophotos
Nigardsbreen
Baklibreen
Bergsetbreen
Max (m d–1)
Average (m d–1)
1.34
2.09
1.61
0.56
0.38
0.53
Table 3. Comparison of GPS measured stake displacements between 22nd June and 22nd August 2001 and between 22nd August and
19th September (Tønsberg 2003), and the average of velocities obtained by cross-correlation of orthophotos acquired 19th and 29th
August 2001 in a 100 m radius circle around the stake positions. The positions of the GPS-measured stakes are shown in Fig. 7
GPS
Point
1
2
3
4
5
6
7
8
Orthophoto
velocity
(m d–1)
GPS velocity
06–08–2001
(m d–1)
0.31
0.61
0.57
0.60
0.62
0.66
0.70
0.77
0.35
0.60
0.59
0.61
0.60
0.67
0.70
0.80
Difference
ortho and GPS
velocity (%)
Average deviation
vertical displacement). A reference window size of
15 × 15 pixels was used for all glaciers. This size
seemed to be sufficient to contain enough information for matching the 0.5 m resolution photos of
glacier surfaces used in this study. The size of the
test area was 100 × 100 pixels. The test area has to
be large enough to detect a displacement of the expected magnitude. The size was determined by
some initial trial testing. According to Kääb and
Vollmer (2000) the accuracy of this method is in the
order of one pixel and at least at the same level as
results from analytical photogrammetry. Although
originally developed for deformation mapping of
rock glaciers using aerial orthophotos, this method
has also been used to map glacier movement using
ASTER satellite data (Kääb 2002, 2005; Kääb et al.
2006).
When a sufficient number of points have been
matched by cross-correlation, it is possible to perform filtering on the resulting data. Displacement
vectors with low cross-correlation factors or unnatural directions can be easily filtered out. A correlation coefficient threshold of 0.8 was used for all locations together with a directional filter, filtering
any vector deviating greatly from the assumed flow
direction. The filtered data were then imported into
a GIS and some manual removal of points was unGeografiska Annaler · 88 A (2006) · 1
–12.9
1.6
–3.5
1.7
3.2
–1.5
0.0
–3.9
3.5
GPS velocity
08–09–2001
(m d–1)
Difference
ortho and GPS
velocity (%)
0.29
6.5
0.53
0.54
0.55
0.60
0.65
0.73
7.0
10.0
11.3
9.1
7.1
5.2
8.0
dertaken. Using a correlation coefficient threshold
of 0.8 and a directional filter adapted to the main
flow direction of the different glaciers as well as
manual editing, we excluded 37–50% of the original matched points.
DEM differencing
Having the 5 m resolution DEMs of 29 August,
2001 and 14 August, 1997, the 25 m resolution
DEMs of 1984 from the Norwegian Water Resources and Energy Directorate (NVE) and of
1993 from the Norwegian Mapping Authority
(Statens Kartverk), the vertical surface changes for
the periods were calculated simply by subtracting
the older model from the newer by the use of ESRI’s Arc GIS software. The 1984 and 1993 DEMs
have been interpolated from contour lines of 10 m
and 20 m equidistance respectively. The 1984 and
1997 DEMs cover only Nigardsbreen. A resampling to 25 m was done for the 2001 and 1997
DEMs in order for them to be comparable with the
others. The height accuracy of the 2001 and 1997
DEMs is assumed to be 0.8 m and 1.3 m (c. 0.025%
of 3100 and 5300 m flying height), while both the
1984 and 1993 DEMs have an accuracy closer to 5
m (Table 1).
59
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 3. Velocity vectors on Nigardsbreen from cross-correlation matching of the 19 and 29 August 2001 orthophotos. The 10 × 10 m
spatial resolution of the original velocity results has been thinned to a 40 × 40 m resolution
Glacier velocities
Velocity estimation
The orthophoto resolution is 0.5 m. The calculated
velocities from cross-correlation matching of orthophotos are shown in Tables 2 and 3. According
to Kääb and Vollmer (2000) the accuracy of the
cross-correlation matching is about the size of one
pixel, giving an accuracy of 0.5 m for the whole
measuring period of ten days (19 and 29 August
2001). This is equivalent to an accuracy of 0.05 m
d–1 when presenting the results from the ten day period as metres per day. The resulting velocities are
presented as vector plots showing the magnitude
60
and direction of the displacements (Figs 3, 4, 5 and
6), and as interpolated velocity fields with velocityisolines (Figs 7, 8 and 9). The interpolation used is
a local second-degree polynomial interpolation.
This method only interpolates the magnitude of the
displacements, and is a smoothing and inexact interpolation method, used to improve visualization
rather than give an exact presentation of the results.
Nigardsbreen
The highest velocities on Nigardsbreen, which
reached 1.34 m d–1, were found in the main icefall
leading down to the valley glacier (Figs 3 and 7).
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 4. An image showing a smaller area of the measured velocity vectors on Nigardsbreen with the velocity results in the original
10 × 10 m resolution
High velocities were also found in the upper part of
the southernmost icefall, where velocities reach
1.19 m d–1. This is in the area where Liestøl measured a displacement of 1.40 m d–1 using terrestrial
photogrammetry in summer 1951 (Østrem et al.
1976) and Bergersen (1954) measured a displacement peak of 0.9 m d–1 in summer 1952. Below the
icefalls velocities gradually decrease towards the
glacier snout. The calculated velocities from crosscorrelation matching of orthophotos are shown in
Table 3. From Figs 3, 4 and 7 it can also be seen that
the velocities of the two western tributaries decrease before they enter the main glacier stream.
Geografiska Annaler · 88 A (2006) · 1
A comparison of the glacier velocities from this
study with the ones measured by GPS between 22
June and 22 August, 2001 and between 22 August
and 19 September, 2001 by Tønsberg (2003), is
shown in Table 3. An average deviation of 3.5% for
the first measurement period and 8.0% for the second is found. The positions of the stakes used for
measurement by Tønsberg (2003) were chosen in
order to be as close as possible to earlier observations. A schematic view of the earlier and new velocity measurements is presented in Table 4. The
average velocity along the centreline of Nigardsbreen from these results was 0.76 m d–1 (276.8 m
61
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 5. Velocity vectors on Baklibreen from cross-correlation matching of the 19 and 29 August 2001 orthophotos pictured in the original
10 × 10 m resolution
a–1). Given the 4.2 km distance from the upper areas covered by velocity measurements to the front,
and the 2.5 km distance further up to the equilibrium line altitude (ELA), a surface travel time of
24.2 years can be calculated from the ELA to the
62
snout for an object (e.g. a stone) travelling on the
glacier surface. We assumed that the average velocity found between the snout and 1100 m a.s.l. is also
applicable for areas all the way up to the ELA.
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 6. Velocity vectors on Bergsetbreen from cross-correlation matching of the 19 and 29 August 2001 orthophotos. The 10 × 10 m
spatial resolution from the original results has been thinned to a 40 × 40 m resolution
Baklibreen
The highest velocities at Baklibreen were found in
the area of the greatest slope, in the lower and thin
part of the glacier, reaching a maximum of 2.09 m
d–1 (Figs 5 and 8, and Table 2). The average velocity
for Baklibreen is 0.38 m d–1. The velocities show
the same pattern of increasing towards the front as
found during the investigations in 1988–1996
(Kjøllmoen 2000). No travel time has been calculated for Baklibreen because the velocity field is
too inhomogeneous and the glacier section covered
by measurements is too small.
from Bergsetbreen for comparison. The following
discussion section outlines possible reasons for the
patchy appearance of velocity patterns. The southern outlet is hidden in shadows in both orthophotos
and it was not possible to calculate velocities for
this part of the glacier. If we assume that Bergsetbreen has an ELA at the same level as Nigardsbreen, the velocity field covers the complete area
from the glacier snout to the ELA at 1500 m a.s.l.
The average velocity along the centreline in these
areas is 256.45 m a–1 and gives an 8.6-year travel
time for the 2.2 km distance.
Bergsetbreen
Bergsetbreen show velocities averaging 0.53 m d–
1 and reaching a maximum of 1.61 m d–1 in the steep
mid-section (Figs 6 and 9, Table 2). On the upper
southern flank and at the flat lower part of the
tongue, the lowest velocities averaged around 0.11
m d–1. No published velocity measurements exist
Surface elevation changes
DEM subtraction
Surface elevation changes were found by subtracting the multi-temporal DEMs (see Table 1 on DEM
data). The total height accuracy of the DEM subtraction and hence the accuracy of the measured
surface elevation changes depend on the accuracy
of each model used and is the Pythagorean sum of
Geografiska Annaler · 88 A (2006) · 1
63
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 7. Velocity field of Nigardsbreen interpolated from the velocity vectors. The location of the GPS points used for velocity measurements by (Tønsberg 2003) is shown
the two DEM accuracies since no correlation is
present. Given the DEM height accuracies in Table
1, the accuracy of each subtraction can be estimated
and is shown in Table 5. These accuracies will have
implications for the interpretation and validity of
the surface elevation change results.
Nigardsbreen
Nigardsbreen showed only minor surface elevation changes from 1984 to 1993 with a thickening
of 5–6 m in the lower parts and surface lowering
64
of up to 10 m in the highest areas (Figs 10 and 11).
The average change for this period was a lowering
of 3.3 m within the 1993 glacier outline (Table 5).
From 1993 to 1997 there was a thickening for the
whole investigated section with an average of 19.3
m within the 1997 glacier outline. (An area in the
middle part had to be masked out due to errors in
the generation of the 1997 DEM; Fig. 10). Due to
an advance of approximately 200 m in this period
the glacier experienced its greatest surface elevation change in the lower areas, with the maximum
value being over 40 m. Above the advancing
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 8. Velocity field of Baklibreen interpolated from the velocity vectors
Geografiska Annaler · 88 A (2006) · 1
65
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 9. Velocity field of Bergsetbreen interpolated from the velocity vectors
snout, thickening was approximately 20 m (Fig.
11). It is also worth noticing the thickening in the
lower part of the northwestern icefall. For the last
period, from 1997 to 2001, the change was somewhat smaller than for the preceding period but still
positive in all the investigated elevation intervals,
showing a rise in the lower advancing areas with
approximately 33 m decreasing to approximately
5 m for the areas above 550 m a.s.l. (Fig. 11). The
average thickening within the 2001 outline was
5.2 m (Table 5). For comparison with Baklibreen
and Bergsetbreen, surface elevation changes were
also calculated for the periods 1984 to 2001 and
1993 to 2001 (Table 5) (The small differences are
due to different outlines being used for different
periods).
Baklibreen
The surface elevation change at Baklibreen from
1993 to 2001 did not show the elevational trend indicated by Nigardsbreen, but showed an average
thickening of 14.3 m with a surface change between –5.6 and +34.8 m (Fig. 12, Table 5).
Table 4. Comparison of glacier velocities (m d–1) conducted by terrestrial photogrammetry in
1937 (Pillewizer 1950), by Liestøl in 1949, 1951, 1953 and 1961 (Østrem et al. 1976), trigonometric stake measurements in 1968/69 (Nielsen 1970), and the GPS and orthophoto measured
ones from 2001 presented here (Tønsberg 2003)
GPS
points
1937
1949
1951
1953
1961
2 and 3
5 and 6
7 and 8
0.32
0.63
0.51
0.32
–
0.73
0.26
–
0.76
0.80
–
0.62
–
66
0.43
1968/
1969
2001
GPS
2001
ortho
0.11
0.37
0.56
0.60
0.64
0.75
0.59
0.64
0.74
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 10. Surface elevation change of Nigardsbreen calculated from differencing the DEMs for 1984–1993, 1993–1997 and 1997–2001
Table 5. Mean surface elevation change (m) from DEM differencing with the mean absolute value
in parentheses. For each calculation the glacier border of the last date has been used. The accuracy
(m) for each surface change calculation is also shown
Nigardsbreen
Baklibreen
Bergsetbreen
Accuracy
Years
1984–2001
1993–2001
1984–1993
1993–1997
22.1
(22.2)
–
24.6
(24.6)
14.3
(14.3)
3.2
(6.6)
5.1
8
–3.3
(5.7)
–
19.3
(19.9)
–
–
–
–
7.1
9
5.1
4
1.5
4
–
5.1
17
Geografiska Annaler · 88 A (2006) · 1
1997–2001
5.2
(5.6)
–
67
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 11. Surface elevation changes
from differencing the DEMs. Also
shown is the area distribution plotted as a function of elevation for
the investigated area on Nigardsbreen. The points making the curve
are average values for zones covering every 50 m of elevation (the
curve to the far right is the area distribution)
Bergsetbreen
Bergsetbreen advanced during the 1993–2001 period and thickened by almost 70 m in its lower parts
but there was only a slight lowering for the rest of
the glacier (Fig. 13). The average surface elevation
change was positive, with a value of 3.2 m (Table
5), although with an accuracy of ±5.1m.
68
Discussion
Glacier velocities
As noted earlier, the accuracy of the cross-correlation matching method is about 0.5 m using orthophotos of 0.5 m resolution, giving accuracy for the
daily velocities of 0.05 m d–1. This means that all
the presented velocities in Table 3 are significant.
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
Fig. 12. Surface elevation change from 1993 to 2001 on Baklibreen from differencing of DEMs
The good agreement between the orthophoto captured velocities at Nigardsbreen and the GPS measured ones from the same period also confirms that
the method of cross-correlation matching is working (Table 3). As seen from Table 3, the similarity
Geografiska Annaler · 88 A (2006) · 1
between GPS-measured velocities is greater in the
June to August period than for the August to September period. This may be explained by a decrease
in surface melting, leading to a drop in subglacial
water pressure and finally resulting in a decrease in
69
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
Fig. 13. Surface elevation change from 1993 to 2001 on Bergsetbreen from differencing of DEMs
sliding velocity in the late summer. The relation between subglacial water pressure and surface velocity is well known (Iken and Bindschadler 1986).
From Table 3 it can also be seen that the greatest
deviation for the orthophoto velocities compared to
the GPS measured ones from June to August was at
point 1. This can be explained partly by the fact that
the accuracy is a constant term having stronger relative effect on the smaller velocities.
Looking at how velocity on the tongue of Nigardsbreen has changed from 1938 to 2001, it is evident that the velocity is heavily affected by changes in glacier thickness, slope and width. The glacier
retreated rapidly until the 1970s and advanced in
the same manner from 1987 until 2003. The glacier
thicknesses of all the earlier measurement periods
in Table 4, except for 1968/69, were greater than
the one in 2001 (Østrem et al. 1976). But given the
70
surface development in Fig. 19 in Østrem et al.
(1976), the surfaces of those periods were also less
steep than the 2001 surface.
Since no detailed information on glacier surfaces
from earlier times is available for direct comparison,
it is difficult to model the impact these differences
have in a detailed manner. However, from Table 4 it
can be seen that the gentle surface slope of the retreating glacier in the late 1960s produced a much
lower velocity at the tongue than the one measured
for the steep front in 2001. The combination of glacier thickness, slope and width is probably also the
explanation for the velocities of points 5 and 6 being
so similar in 1937 and 2001, and the similarity in velocities of points 7 and 8 in 1949, 1951 and 2001.
This means that the greater slope counteracted the
effect of a thinner glacier in 2001, leaving the velocity more or less the same as in 1937 and 1949/50.
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
The reasons for the lower velocities at the lower
part of two western icefalls can also be attributed to
glacier slope and thickness. The glaciers are probably rather thin in both of these icefalls, and when
the slopes become gentler just before they enter the
main stream, the velocities decrease. This is best
seen in Fig. 4 where there is a huge contrast in the
velocities between the main stream and the tributary. This will also have implications for how the
glacier geometry responds to a mass balance
change.
There are no contemporary ground measurements of velocity at Baklibreen or Bergsetbreen.
For Baklibreen there has been an increase in velocity since the monitoring project between 1987 and
1999 (Kjøllmoen 2000). This increase can also be
attributed to a change in glacier thickness as shown
in Fig. 12. Judging from the orthophotos Baklibreen is very thin and partly disintegrating in its
steep lower parts, this indicates sliding rather than
deformation as the main velocity component. It is
in this hanging part of the glacier that ice avalanches usually originate. It is also in this very steep area,
with slopes of 30° inclination, that we find the
greatest velocities for Baklibreen and for all the
three glaciers investigated.
As stated before, the velocity field of Bergsetbreen (Fig. 9) has a patchy appearance with the velocity changing rapidly within relatively small areas. Bergsetbreen is probably a rather thin glacier
in its steep icefall, meaning that small changes in
bottom topography will have an effect on the velocities seen on the surface. From the velocity field
it can also be seen that there is a major difference
in the velocity of the main drainage channel compared to the southern flank. The decreasing velocity
towards the tongue is also due to changes in surface
slope. The velocity pattern of the tributary icefalls
on Nigardsbreen is also a property that would affect
how glacier geometry responds to mass balance
change.
Surface elevation changes
If one compares the mean surface changes for each
glacier and period with the accuracy of the different
surface change calculations (Table 5), it is clear that
the overall change at Bergsetbreen from 1993 to
2001 and at Nigardsbreen from 1984 to 1993 is not
significant; the same is found; when considering
the mean of the absolute value of the changes (Table 5). Nevertheless, there seems to be an altitudinal trend with a thickening in the lower parts and a
Geografiska Annaler · 88 A (2006) · 1
thinning in the upper parts for Nigardsbreen from
1984 to 1993 (Fig. 11), but due to the accuracy this
interpretation should be treated with great care.
Looking at Fig. 13 it is clear that even though the
overall surface elevation change at Bergsetbreen
was not significant, there is a significant thickening
at the advancing front, with vertical growth reaching almost 70 m. Based on these data this means
that Bergsetbreen’s surface has been stable from
1993 to 2001 for all areas except for the advancing
front, where there is a significant thickening. The
reason for a stable surface elevation in the icefall is
probably due to the steepness of Bergsetbreen. This
steepness made it possible for the mass balance surplus from the late 1980s and early 1990s to be rapidly transported down the steep and thin icefall,
giving a short reaction time for frontal changes below the icefall.
Baklibreen showed a thickening of 12.9 m from
1993 to 2001, and also a great spatial variability in
the surface elevation increase, varying between –
5.6 and 34.8 m. The great variance at Baklibreen
can be explained by its rugged and partly disintegrated surface. The thin ice cover will also show
huge differences due to small horizontal displacements in the disintegrated areas that cannot be attributed to changes in the total volume of the glacier. For comparison, the surface change data from
the earlier works (Kjøllmoen 2000) showed an increase in ice thickness of 10 to 20 m from 1989 to
1994, a slight increase from 1994 to 1996 and no
visible changes from 1996 to 1999. Measurements
in 2001 showed a lowering of 0 to 3 m from 1999
(Kjøllmoen 2003a). Nevertheless, it seems that the
increase revealed in this study from 1993 to 2001
is mainly explained by a thickening in the first few
years of the period. One should also bear in mind
that the investigations are not done for exactly the
same area.
Having four available DEMs for Nigardsbreen
made it possible to study the surface elevation
changes in a more detailed manner. In general, Nigardsbreen was stable from 1984 to 1993 (there
was a non-significant average surface lowering of
3.3 m), but experienced an average thickening of
19.3 m from 1993 to 1997, and an average thickening of 5.2 m from 1997 to 2001. From Fig. 11 it
can also be seen that there were altitudinal differences for all three periods. The most striking feature is the great thickening near the snout in the two
latter periods caused by the advancing front. There
are also small signs of an advance in the 1984–1993
data with an average thickening of 7.1 m in the ar71
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
eas around 525 m a.s.l. Since it is the ablation area
that is investigated, the surface elevation change
must be due to increasing mass transport from
higher elevations.
At the lower part of the northwestern icefall a
thickening of up to 58 m for the 1993 to 1997 period
can be seen (Fig. 10). Looking at the velocity pattern in this area, it can be seen that velocity decreases with decreasing altitude (Fig. 7). Mass surplus
transported down this icefall would, as has been
discussed for Bergsetbreen, lead to a mass build-up
where the velocity decreases at the bottom of the
icefall, because the high velocity of the main glacier stream would obstruct the tributary from rapidly deploying its mass further down.
Calculated transport times from the equilibrium
line to the terminus for Nigardsbreen and Bergsetbreen were 24.2 and 8.6 years, respectively. According to Nye (1960), a kinematic wave travels
three to five times faster than the surface velocity.
According to van de Wal and Oerlemans (1995) one
would usually not detect terminus changes when a
kinematic wave arrives because of diffusion. But
for steep glaciers like Nigardsbreen (10–20°) and
Bergsetbreen (27°) the diffusion term would be
very small because of its inverse dependence of
slope, and the travel times for kinematic waves described by Nye (1960) valid.
The steepness of the two glaciers would also
lead to basal sliding being a substantial part of the
total surface velocity. Estimating deformation profiles (Paterson 1994) based on DEM slope and the
few thickness data sets that exist reveals that basal
sliding probably accounts for at least 40–60% of
the total surface velocity on Nigardsbreen. Kinematic wave velocity is a weighted sum of glacier
deformation (weight 5) and sliding (weight 3). The
proportion of basal sliding on Nigardsbreen implies that the kinematic wave should not travel
more than four times faster than the surface velocity. For Nigardsbreen and Bergsetbreen this means
that it would take around six and two years, respectively, for a kinematic wave to travel from the ELA
down to the front of the glacier.
Comparing mass balance data (Kjøllmoen
2003b) with length change data (Kjøllmoen 2001)
for Nigardsbreen, one sees that the great mass balance surplus which started in 1987 (1.48 m water
equivalents (w.e.)) and continued in 1989 and
1990 (3.2 and 1.77 m w.e.) gave the greatest front
changes in the years 1994 to 1996 (36, 50 and 40
m change). There were also advances of 10, 21 and
14 m in the years 1991, 1992 and 1993 (Kjøllmoen
72
2001). This implies that there was a five to seven
year lag from the abrupt changes in mass balance
until the glacier advanced.
If one regards the abrupt change in mass balance
in the late 1980s as a perturbation that could initiate
a kinematic wave, the time lag for changes at the
front agree quite well with Nye’s theory for kinematic waves. It should be noted, however, that several assumptions have been made on basal sliding
and lack of diffusion etc. Nevertheless, it shows that
both Nigardsbreen and Bergsetbreen are glaciers
that react fast to climatic changes due to their high
velocity and steep gradient. For Nigardsbreen this is
amplified by having a large accumulation area
draining into a narrow tongue (Oerlemans 1992).
Assuming that Nigardsbreen and Bergsetbreen
experienced the same changes in mass balance during the late 1980s and early 1990s, it appears that
they reacted in somewhat different geometrical
ways. From 1993 to 2001 Bergsetbreen only
showed significant changes in its lower parts (Fig.
13) while Nigardsbreen showed a general thickening in the whole ablation area (Fig. 11). Both
showed thickening at the front and were advancing.
Even though travel times of kinematic waves cannot be taken as expressions of the time it takes to
transport the mass surplus itself, they may reflect
differences in this capability. The steep icefall of
Bergsetbreen would be able to transport a mass balance surplus more rapidly down to the terminus
than Nigardsbreen. This explains why we only see
a frontal change for Bergsetbreen whereas we see
a change for the whole investigated area at Nigardsbreen.
Conclusion
The results from this work show how high-resolution measurements from cross-correlation matching of orthophotos can be used to assess the velocity field of a glacier over a few days. The measured
velocities for three different outlets of Jostedalsbreen are in good agreement with GPS-measured
velocities for Nigardsbreen during the same time
period. The average velocities ranged from 0.38 to
0.56 m d–1 for the three glaciers Baklibreen, Bergsetbreen and Nigardsbreen. Based on the differences between DEMs from 2001 and 1993 on all glaciers, in addition to DEMs from 1997 and 1984 at
Nigardsbreen, an average increase in surface elevation of 22.1 m from 1984 to 2001 was found for
Nigardsbreen and an increase of 3.2 and 14.3 m for
Bergsetbreen and Baklibreen for the period of 1993
Geografiska Annaler · 88 A (2006) · 1
SURFACE VELOCITIES OF OUTLET GLACIERS
to 2001. From the surface elevation change it can be
seen how the three different glaciers reacted to the
highly positive winter balance of the late 1980s and
early 1990s. These differences can be explained by
differences in slope, glacier thickness, hypsometry
and velocity.
Acknowledgements
The 2001 air photo campaign which made this
study possible was financed by the Norwegian Water Resources and Energy Directorate (NVE), and
the Department of Geosciences, University of
Oslo, through the EU project Glaciorisk (EVG1–
2000–00512). Thanks are extended to the NVE for
providing the 1984 DEM and to Bjarne Kjøllmoen
at NVE for providing details on accuracy for the
same model. We would also like to thank Jan Harald Tallhaug at the Norwegian Mapping Authority
for information on the quality and photo dates of
the 1993 DEM provided. Kjetil Melvold, Gaute
Lappegard and Thomas Schuler provided valuable
discussions on some of the topics addressed in this
paper. Additional thanks are extended to Kari Anita
Pulver for improving the English.
Bjørn Wangensteen, Department of Geosciences,
University of Oslo, PO Box 1047 Blindern, NO0316 Oslo, Norway.
Ole Magnus Tønsberg, Department of Geosciences, University of Oslo, PO Box 1047 Blindern, NO0316 Oslo, Norway
Andreas Kääb, Department of Geography, University of Zürich-Irchel, Winterthurerstrasse 190, CH8057 Zürich, Switzerland
Trond Eiken, Department of Geosciences, University of Oslo, PO Box 1047 Blindern, NO-0316
Oslo, Norway
Jon Ove Hagen, Department of Geosciences, University of Oslo, PO Box 1047 Blindern, NO-0316
Oslo, Norway
References
Andreassen, L.M., Elvehøy, H., Kjøllmoen, B., Engeset, R.V. and
Haakensen, N., 2006: Glacier mass balance and length variation in Norway. Annals of Glaciology, (in press).
Baltsavias, E.P., 1996: Digital ortho-images – a powerful tool for
the extraction of spatial- and geo-information. ISPRS Journal
of Photogrammetry and Remote Sensing, 51: 63–77.
Geografiska Annaler · 88 A (2006) · 1
Bergersen, A., 1954: En undersøkelse av Svartisen ved Holandsfjorden 1950–51, med tillegg: Hastighetsmålinger ved Nigardsbreen, Jostedalen 1952 (in Norwegian). MSc thesis. Department of Physical Geography, University of Oslo, Norway.
Bindschadler, R.A. and Scambos, T.A., 1991: Satellite-image-derived velocity field of an Antarctic ice stream. Science,
252(5003): 357–364.
Delacourt, C., Allemand, P., Casson, B. and Vadon, H., 2004: Velocity field of ‘La Clapière’ landslide measured by the correlation of aerial and QuickBird satellite images. Geophysical
Research Letters, 31(15).
Finsterwalder, R., 1931: Geschwindigkeitsmessungen an Gletschern mittels Photogrammetrie. Zeitschrift für Gletscherkunde für Eiszeitforschung und Geschichte des Klimas,
19: 251–262.
Haeberli, W., King, L. and Flotron, A., 1979: Surface movement
and lichen cover studies at the active rock glacier near the Grubengletcher, Wallis, Swiss Alps. Arctic and Alpine Research,
11(4): 421–441.
IAHS(ICSI), UNEP and UNESCO, 1998: Fluctuations of glaciers
1990–1995. In: Haeberli, W., Hoelzle, M., Suter, S. and Frauenfelder, R. (eds): Vol. 7. International Association of Hydrologic Sciences. Paris. 296 p.
IAHS(ICSI), UNEP, UNESCO and WMO, 2001: Glacier Mass
Balance Bulletin No. 6. In: Haeberli, W., Frauenfelder, R. and
Hoelzle, M. (eds): World Glacier Monitoring Service. Zürich.
93 p.
Iken, A. and Bindschadler, R.A., 1986: Combined measurements
of subglacial water pressure and surface velocity of Findelengletscher, Switzerland: Conclusions about drainage system av
sliding mechanism. Journal of Glaciology, 32(110): 101–
117.
Kasser, M. and Egels, Y., 2002: Digital Photogrammetry. Taylor
and Francis. London, 351 p.
Kaufmann, V. and Ladstädter, R., 2003: Quantitative analysis of
rock glacier creep by means of digital photogrammetry using
multi-temporal aerial photographs: Two case studies in the
Austrian Alps. In: Phillips, M., Springman, S.M. and Arenson, L.U. (eds): Permafrost: Proceedings of the Eighth International Conference on Permafrost, Zurich, Switzerland, 21–
25 July 2003. 1: 525–530.
Kjøllmoen, B. (ed.) 2000: Glasiologiske undersøkelser i Norge –
Glaciological investigations Norway 1999. Rapport nr 2.
Norges Vassdrags- og Energidirektorat (NVE). Oslo. Norway. 140 p.
Kjøllmoen, B. (ed.) 2001: Glaciological investigations in Norway
in 2000. Report no 2. Norwegian Water Resources and Energy
Directorate (NVE). Oslo. Norway. 122 p.
Kjøllmoen, B. (ed.) 2003a: Glaciological investigations in Norway in 2001, Report no 1. Norwegian Water Resources and
Energy Directorate (NVE). Oslo. Norway. 102 p.
Kjøllmoen, B. (ed.) 2003b: Glaciological investigations in Norway in 2002. Report no 3. Norwegian Water Resources and
Energy Directorate (NVE). Oslo. Norway. 92 p.
Kjøllmoen, B. (ed.) 2004: Glaciological investigations in Norway
in 2003, Report no 4, Norwegian Water Resources and Energy
Directorate (NVE). Oslo. Norway. 98 p.
Kääb, A., 2002: Monitoring high-mountain terrain deformation
from repeated air- and spaceborne optical data: examples using digital aerial imagery and ASTER data. ISPRS Journal of
Photogrammetry and Remote Sensing, 57(1–2): 39–52.
Kääb, A., 2005: Combination of SRTM3 and repeat ASTER data
for deriving alpine glacier flow velocities in the Bhutan Himalaya. Remote Sensing of Environment, 94(4): 463–474.
Kääb, A. and Funk, M., 1999: Modelling mass balance using photogrammetric and geophysical data: a pilot study at Gries-
73
BJØRN WANGENSTEEN, OLE MAGNUS TØNSBERG, ANDREAS KÄÄB, TROND EIKEN AND JON OVE HAGEN
gletcher, Swiss Alps. Journal of Glaciology, 45(151): 575–
583.
Kääb, A. and Vollmer, M., 2000: Surface geometry, thickness
changes and flow fields on creeping mountain permafrost: automatic extraction by digital image analysis. Permafrost and
Periglacial Processes, 11: 315–326.
Kääb, A., Haeberli, W. and Gudmundsson, G.H., 1997: Analysing
the creep of mountain permafrost using high precision aerial
photogrammetry; 25 years of monitoring Gruben rock glacier,
Swiss Alps. Permafrost and Periglacial Processes, 8(4):
409–426.
Kääb, A., Lefauconnier, B. and Melvold, K., 2006: Flow field of
Kronebreen, Svalbard, using repeated Landsat7 and ASTER
data. Annals of Glaciology, (in press).
Nielsen, C., 1970: Triangulering av bevegelsesretninger i 1969.
In: Pytte, R. (ed.): Glasiologiske undersøkelser i Norge 1969.
Hydrologisk avdeling. Norges Vassdrags og elektrisitetsvesen. 96 p.
Nye, J.F., 1960: The response of glaciers and ice sheets to seasonal and climatic changes. Proceedings of the Royal Society
of London, Series A, 257(1216): 589–584.
Oerlemans, J., 1992: Climate sensitivity of glaciers in southern
Norway: application of an energy-balance model to Nigardbreen, Hellstugubreen and Ålfotbreen. Journal of Glaciology,
38(129): 223–232.
Paterson, W.S.B., 1994: The Physics of Glaciers. Pergamon/Elsevier Science. Oxford. 480 p.
74
Pillewizer, W., 1950: Bewegungsstudien an Gletschern des Jostedalsbre in Süd-Norwegen. Erdkunde, 4: 201–206.
Rolstad, C., 1995: Satellitt- og flybilder til bestemmelse av bredynamikk (in Norwegian). MSc thesis. Departement of Physical Geography, University of Oslo. Norway. 129 p.
Tønsberg, O.M., 2003: Kartlegging av bredynamikken på en
fremrykkende bre -Nigardsbreen, Sør-Norge – vha GPS og
fotogrammetri (in Norwegian). MSc thesis. Departement of
Physical Geography, University of Oslo. 65 p.
van de Wal, R.S.W. and Oerlemans, J., 1995: Response of valley
glaciers to climate change and kinematic waves: as study with
numerical ice-flow model. Journal of Glaciology, 41(137):
142–152.
Østrem, G., Liestøl, O. and Wold, B., 1976: Glaciological investigations at Nigardsbreen, Norway. Norsk Geografisk Tidskrift – Norwegian Journal of Geography, 30(1): 187–209.
Østrem, G., Dale Selvig, K. and Tandberg, K., 1988: Atlas over
breer i Sør-Norge – Atlas of Glaciers in South Norway. Meddelelse nr. 61 fra Hydrologisk Avdeling. Norges Vassdragsog Energiverk. 248 p.
Østrem, G., Haakensen, N., Kjøllmoen, B., Laumann, T. and
Wold, B., 1991: Massebalansemålinger på norske breer 1988
og 1989. NVE-Publikasjon nr 11. Hydrologisk avdeling, Norges vassdrags- og energiverk. 78 p.
Manuscript received June 2005, revised and accepted November
2005.
Geografiska Annaler · 88 A (2006) · 1
Norsk Geografisk Tidsskrift–Norwegian Journal of Geography Vol. 59, 276–285. Oslo. ISSN 0029-1951
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data
BJØRN WANGENSTEEN, DAN JOHAN WEYDAHL & JON OVE HAGEN
Wangensteen, B., Weydahl, D.J. & Hagen, J.O. 2005. Mapping glacier velocities on Svalbard using ERS tandem DInSAR data.
Norsk Geografisk Tidsskrift–Norwegian Journal of Geography Vol. 59, 276–285. Oslo. ISSN 0029-1951.
The results presented show how glacier velocities can be measured and calculated from Earth Resources Satellite (ERS) tandem
interferometric synthetic aperture radar (InSAR) data by the use of geographical information systems (GIS) and noninterferometric software. A semi-automatic algorithm using existing and newly implemented modules in ESRI’s GIS software Arc
has been developed to unwrap existing height differentiated interferometric SAR (DInSAR) data. The algorithm is applied to a data
set covering glaciers from the arctic archipelago Svalbard. Estimated DInSAR velocities are decomposed into the glacier surface
flow direction using a digital elevation model (DEM). Velocity fields for the glaciers Isachsenfonna, Akademikerbreen and
Nordbreen on Svalbard are presented.
Keywords: glaciology, remote sensing, satellite SAR interferometry, Svalbard
Bjørn Wangensteen, Jon Ove Hagen, Department of Geosciences, University of Oslo, P.O. Box 1047 Blindern, NO-0316 Oslo,
Norway. Dan Johan Weydahl, Norwegian Defence Research Establishment (FFI), P.O. Box 25, NO-2027 Kjeller, Norway.
E-mail: bjorn.wangensteen@geo.uio.no
Introduction
Detailed knowledge of glacier velocity fields is important to
increase knowledge of glacier dynamics and to verify models
dealing with this topic. Glacier surface velocities are also
an important part of mass balance modelling of glaciers and
therefore constitute an important parameter for monitoring
how glaciers respond to a changing climate. Many of
Svalbard’s glaciers are of the surge type, which makes
glacier velocity monitoring even more important. Estimates
of the percentage of surge-type glaciers vary from 15 to 90
(Liestøl 1969, Lefauconnier & Hagen 1991, Hagen et al.
1993, Hamilton & Dowdeswell 1996, Jiskoot et al. 2000).
Satellite remote sensing of glacier velocities is less expensive
and covers larger areas than traditional surveying in remote
areas such as the Arctic. Other remote sensing techniques,
such as feature tracking in optical satellite imagery or aerial
photos, require areas of high contrast in order to work
(Lefauconnier et al. 1994, Kääb & Funk 1999). Hence,
feature tracking only works in crevassed or debris-covered
areas on glaciers, and serves as a complimentary technique
to InSAR, which works best in more homogenous and
stable areas.
Glacier surface velocities were first measured with data
from satelliteborne InSAR in 1993 (Goldstein et al. 1993).
The InSAR technique has been extensively used for mapping
glacier surface velocity; see Massonnet & Feigl (1998) and
König et al. (2001) for an overview. The launch of ERS-2 in
1995 made available tandem scenes from ERS-1 and ERS-2
taken only one day apart. The short time interval between
acquisitions of the SAR scenes increases the possibility for
detection and mapping of glacier movements as there is less
influence from temporal factors such as melting. The scenes
used in this study are from the period of the tandem mode in
1995 and 1996.
The main goals of this study were to develop a method
for unwrapping existing DInSAR data of three Svalbard
glaciers, to estimate glacier velocities from the unwrapped
DInSAR data, and to decompose these velocities into the
direction of glacier surface flow using a photogrametrically derived DEM. All these tasks were performed using the
ESRI GIS software Arc and no interferometry software was
used.
The glaciers investigated in this study are Isachsenfonna,
Nordbreen and Akademikerbreen (Fig. 1). Isachsenfonna is
the only one of these for which previous velocity data exists
and the InSAR scenes of this glacier were therefore chosen
to develop the method.
Interferometric SAR
InSAR uses the phase information in radar acquisitions from
two receiving antennas to produce an interferogram (InSARscene). The two antennas can be separated in either time
(repeat track acquisition) or space (along- or across-track).
An interferogram is a 2-dimensional representation of the
measured radar phase difference between the two antenna
positions and corresponding positions on the Earth’s surface.
Several factors contribute to the phase difference, with
ground surface topography and ground surface deformation
being the two most important.
Geometry
The geometry of a satellite InSAR acquisition is shown in
Fig. 2. The first SAR scene is acquired in orbit O1 and the
second in orbit O2. These two orbits are separated by the
spatial baseline B, which can be decomposed into Bx and By.
The difference in distance from a point on the ground to the
two antennas (r2 and r1) can be expressed as a certain number
of wavelengths and a fraction of a wavelength: a phase
difference. This is the phase difference depicted in the
DOI 10.1080/00291950500375500 # 2005 Taylor & Francis Group Ltd
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data 277
O2
B
By
O1
B
Bn
Bx
r2
r1
Fig. 2. The geometry (side view) of an InSAR-acquisition. First acquisition
for the InSAR-scene in O1 and second in O2. B is the spatial baseline between
the two satellite orbits.
Differential interferometry
Fig. 1. Map of Svalbard, showing the location of the three glaciers:
(1) Isachsenfonna, (2) Nordbreen, (3) Akademikerbreen, Longyearbyen, the
research settlement Ny-Ålesund in Kongsfjorden, Åsgårdsfonna (Å),
Wijdefjorden (W), Negribreen (N), and Lomonosovfonna (L).
Satellite InSAR scenes are usually based on two scenes
acquired some time apart (temporal baseline). If surface
deformation has taken place during this time interval it will
also cause phase differences. The phase difference as noted
in Eq. 1 could therefore be expanded with a second part for
the deformation (Kwok & Fahnestock 1996):
Df =
InSAR scene and can be written as the following (Gens &
van Genderen 1996):
4p
4p
(r2 r1 ) =
(Bx sin y By cos y)
Df =
l
l
(Eq: 1)
Here, y is the look angle and l the wavelength (Fig. 2).
Unwrapping
The phase difference is measured modulo 2p; in intervals
from 0 to 2p. It is the shifts from 2p to 0, or 255 to 0 in terms
of grey values, that create the characteristic fringe pattern.
Converted to wavelengths, one fringe equals half a wavelength due to the double pathway. The fringes represent
ambiguities and should therefore be unwrapped. The
principal method of all-phase unwrapping is that the phase
differences are integrated by adding multiples of 2p when
a fringe transition is crossed (i.e. when the phase difference
wraps from 2p to 0). A more thorough description of the
different unwrapping methods and algorithms is given in
Gens & van Genderen (1996). In principle, after unwrapping,
it is possible to create a DEM by using one known elevation
point together with a given spatial baseline to deduct the
elevations for every pixel in the whole scene.
4p
4p
(Bx sin y By cos y) þ
Dr
l
l
(Eq: 2)
where Dr is the deformation component in range during
the temporal baseline. As can be seen from Eq. 2, the
contribution from the deformation is only dependant on
the temporal baseline and not the spatial baseline. In an
interferogram where deformation has taken place one has
to isolate the contribution from one of the two components
topography or deformation to utilize the other. This is
called differential interferometry or DInSAR and can
be done by using two interferograms (three SAR-scenes)
(Kwok & Fahnestock 1996) or one interferogram and a
DEM (Eldhuset et al. 2003).
Limitations
A suitable interferogram can only be obtained if the two
signals received from the same object are coherent (i.e. not
decorrelated). Decorrelation of the spatial baseline is a factor
that is governed by acquisition parameters, and can be
avoided by choosing image pairs with a spatial baseline
shorter than the critical spatial baseline, and for ERS-1 and
ERS-2 this is c.1.1 km.
Another effect of the spatial baseline length is that shorter
baselines will result in less contribution from topography.
There are examples of applying interferograms with spatial
278 B. Wangensteen, D.J. Weydahl & J.O. Hagen
baselines of only a few meters or for nearly flat ice-sheet
surfaces, and thereby completely disregarding the topographic effect (Goldstein et al. 1993, Michel & Rignot 1999).
Temporal effects are changes that occur at the Earth’s
surface during the time between the acquisitions and these
will also have an impact on the coherence of the interferogram. These could be changes in the reflective properties
of the material at the surface. Changes in the water content
will drastically alter the dielectric properties of the ground
and, hence, also the backscatter of the electromagnetic
waves. Motion on the Earth’s surface could also decorrelate
the image if it is too chaotic, having a discrete rather than a
continuous fashion. Glaciers moving with high velocities
(several metres per day) and that are heavily crevassed will
often decorrelate (Weydahl 2001).
Mapping glacier velocities by satellite SAR
interferometry
Goldstein et al. (1993) demonstrated the possibilities of
satellite SAR interferometry for mapping the glacier velocity
of an Antarctic ice stream. The technique has been widely
used for mapping glacier movement on Greenland and
Svalbard, and in the Alps (Joughin et al. 1995, Eldhuset
et al. 1996, Rott & Siegel 1996, Unwin & Wingham 1997).
All of these studies have only measured the component
of the glacial motion in the direction towards the satellite,
also called the line of sight component. Traditional ground
measurements of glacier velocity are done in the direction
of movement. Rignot et al. (1995) used an existing altimetric
DEM to compute slope gradient and slope direction in
Western Greenland. They also made the normal assumption
of glacier movement being surface parallel and calculated
velocities that were within 6% of ground measurements.
Kwok & Fahnestock (1996) used two interferograms to
separate motion and topography. The resulting DEM was
used to decompose the glacier movement into the direction
of slope.
When measuring movement with InSAR data, only the
component of motion in the slant range (i.e. radar beam or
line of sight) direction is measured. If all motion is taking
place in the azimuth direction, there will be no component in
slant range, and hence no motion-induced contribution to the
interferogram. Mohr et al. (1998) combined interferograms
from ascending and descending orbits to map an ice stream
on Greenland in three dimensions, thereby omitting the
problem of measuring motion in the azimuth direction. They
also assumed surface parallel flow.
Study sites
All the glaciers used in this study are situated at Spitsbergen,
the largest island of the arctic archipelago Svalbard (Fig. 1).
Svalbard is situated between 74 N–84 N and 10 E–35 E in
the North Atlantic, and has permafrost conditions down to
sea level (Liestøl 1977). Glaciers of different types cover
60% of the archipelago.
All the following data about the glaciers are taken from
Glacier Atlas of Svalbard and Jan Mayen (Hagen et al.
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
1993). The method was developed using a DInSAR scene
of Isachsenfonna (78 850 N, 13 110 E), shown as site 1 in
Fig. 1. This glacier is situated close to Kongsfjorden and the
research settlement in Ny-Ålesund. Isachsenfonna is classified as an outlet from an ice cap with a composite firn area
and a fast-retreating calving front. The equilibrium line
altitude (ELA) is c.670 m a.s.l. The ELA is in the lower part
of the area where the velocity mapping was performed.
Isachsenfonna merges with Holtedahlsfonna and is called
Kronebreen (the latter two are not named on Fig. 1) further
down.
The second glacier, Nordbreen (‘2’ in Fig. 1), is
considerably smaller than the other two. It is situated in
the north-eastern part of Spitsbergen (79 380 N, 16 100 E).
Nordbreen is an outlet from the Åsgårdsfonna ice cap, with
one firn area and a calving front in Wijdefjorden. The front
is probably retreating slowly. The ELA is c.450 m a.s.l.
For site 3 in Fig. 1, the name Akademikerbreen is used as a
joint name for Transparentbreen, Opalbreen (the latter two
are not named on Fig. 1), the lower part of Akademikerbreen
and parts of Negribreen and Lomonosovfonna. Since
Akademikerbreen occupies the greater part of the InSAR
scene, this name is used for reference purposes. The centre of
the scene is situated in the eastern part of Spitsbergen
(78 400 N, 18 400 E) and the glacier is classified as an outlet
from an ice cap with more than one firn area and a fastretreating calving front. The area of the whole Negribreen
glacier system is 1180 km2, the length is 41 km and the
volume 250 km3. The ELA is c.360 m a.s.l.
Due to their position close to the research settlement in
Ny-Ålesund, the glaciers in the Kongsfjorden area are among
those that have been most investigated on Svalbard. The
glaciers of site 1 are situated within this area. The lower part
of Kronebreen, which drains Isachsenfonna, is heavily
crevassed and moves at a very high velocity. Due to this,
the interferograms decorrelate in this area. Using SPOTimagery, Lefauconnier et al. (1994) found velocities between
0 and 2.15 m d1 on the tongue of Kronebreen. Maximum
summer velocities of 4.5 m d1 have been measured at the
very front (Voigt 1965, Lefauconnier et al. 1994) and this is
much higher than for other similar glaciers on Svalbard.
However, the high velocity is not sufficiently high compared
with the calving rate and the front has therefore been
retreating over the last 200 years. Due to the high velocity,
the glacier is not believed to be frozen to its bed (Liestøl
1988). The velocity measurements referred above were done
below the areas where velocities are mapped in this study.
Lefauconnier et al. (2001) have reported on a GPSmeasured displacement for three stakes at Holtedahlsfonna
from May 1996 to May 1997 that corresponds to 0.10
to 0.15 m d1. Unfortunately, the stakes are situated just
outside the area mapped in this project. Lefauconnier et al.
(2001) also report on nearly constant velocities ranging
from 0.20 to 0.22 m d1 for Isachsenfonna, measured with
satellite SAR interferometry from ERS-1 scenes acquired
six and nine days apart in September and October 1991.
Velocities ranging from 0.08 5 to 0.52 m d1 were obtained
from the same ERS-1 scenes for Holtedahlsfonna, having the
highest velocities in the lower part were the Holtedahlsfonna
becomes Kronebreen.
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data 279
Table 1. Data on the InSAR-scenes used in the study. Bn is the orthogonal satellite baseline in metres.
Glacier
Date (ERS1/ERS2)
Orbit (ERS1/ERS2)
Frame
Track
Bn (m)
Isachsenfonna
Isachsenfonna
Akademikerbreen
Nordbreen
Nordbreen
Nordbreen
10/11.10.96
05/06.04.96
20/21.10.95
27/28.09.95
05/06.04.96
10/11.05.96
23472/3799
24703/5030
22298/2625
21969/2296
24703/5030
25204/5531
1989
1989
1989
1971
1971
1971
438
166
266
438
166
166
31
10
22
130
12
68
Data set
From the launch of ERS-2 in 1995 until ERS-1 ended
its eight-year service in 1999 it was possible to obtain
InSAR-scenes combined from the two satellites. Both
satellites have a repeat cycle of 35 days. During the ninemonth tandem mode mission from August 1995 to May
1996, the satellites had the same orbit and repeat cycle, but
followed each other one day apart. It is data from this tandem
period that is used in this study.
The InSAR scenes were processed at the Norwegian
Defence Research Establishment (FFI) with data from
an ESA AO (European Space Agency Announcement of
Opportunity) Tandem project. Using a DEM delivered by the
Norwegian Polar Institute, the topographic influence was
removed during the processing at FFI to produce height
differentiated InSAR scenes (DInSAR). The method is more
thoroughly described by Eldhuset et al. (2003). This DEM
was also used in the decomposition of glacier velocities.
The DEM is interpolated from photogrammetric data with
an equidistance of 50 m. The relative error of the DEM is
believed to be c.10 m for glacier surfaces (see Discussion).
Implications of this error on the resulting velocities will be
discussed later in this article. The DInSAR scenes used
in this study were chosen from a larger number of scenes
available. Scenes from several glaciers were chosen in order
to be able to test the method on more than one glacier. The
InSAR-scenes used in the study are listed in Table 1. The
spatial baseline Bn corresponds to the orthogonal baseline
defined by the European Space Agency (ESA) and shown in
Fig. 2. ESA is also the source of the baseline data shown
in Table 1 (http://odisseo.esrin.esa.it; accessed September
2005). Spatial baselines with decimetre accuracy were used
during the InSAR processing at FFI. These precise orbit
(PRC) data were obtained from ESA.
The ground resolution of the three first scenes in Table 1 is
16 m20 m. The three last scenes were processed in a
different way and have a ground resolution of 64 m80 m.
Since both the DInSAR-scenes and the DEM are in SARgeometry, all the calculations were done in this geometry.
Georeferencing and transformation to a map projection was
not done since it was not necessary in order to demonstrate
the method.
Methods
The method for calculating glacier velocities was developed
as a semi-automatic algorithm using DInSAR-scenes of
Isachsenfonna. The method was then applied to the rest of the
InSAR-scenes. The flow chart of the method is shown in
Fig. 3. All of the steps have been done using the GIS package
ARC (Version 8.2).
In order to unwrap the DInSAR-scene, the discrete fringe
transitions had to be identified. First, a 33 median filter is
applied twice to reduce the characteristic speckle noise that is
present in SAR and InSAR images. The result is a smoother
InSAR image with the fringe transitions still intact. The raw
and median-filtered DInSAR are shown in Fig. 4a and 4b
respectively. The scenes from Nordbreen have been processed with 16-looks during the SAR processing at FFI,
rather than 4-looks (as for the other scenes). This reduces the
speckle noise considerably (at the cost of spatial resolution),
and the noise level is consequently also reduced in the
interferograms. It was therefore unnecessary to apply the
median filter to the Nordbreen interferograms.
An edge-detecting filter is applied after the median
filtering. It is a derivative filter that uses the Prewitt-operator
InSAR
(elevation differentiated)
Median filtering
Gradient filtering
Thresholding
Vectorizing, topology
+
Phase unwrapping
DEM
+
Velocity calculation
Fig. 3. Flow scheme of the method for calculating glacier velocities from the
differential InSAR scenes.
280 B. Wangensteen, D.J. Weydahl & J.O. Hagen
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Fig. 4. (a) Raw DInSAR, (b) median filtered DInSAR, (c) gradient filtered DInSAR, (d) thresholded DInSAR, (e) vectorized fringe lines and (f) unwrapped
DInSAR. All images are from the 10–11 January 1996 interferogram of Isachsenfonna. Azimuth is in the direction of left to right, and range is from top to bottom
in the images.
to calculate the gradient of the pixel values in the directions
of azimuth and range. The gradient filtered image is shown
in Fig. 4c. Lin et al. (1994) also used the Prewitt-operator
in addition to two similar masks to detect lines at an angle
of 45 . Since the Prewitt-operator implemented in Arc gave
satisfying results for the 45 lines, further detection of these
was not pursued.
There is still some noise in the image after the Prewittfiltering and the image is therefore thresholded. The user
can define the threshold after looking at the histogram of
the image. All pixels with values less than the threshold are
set to 0, and the others are kept as they are. A threshold
of c.150 has been shown to be suitable. Determining the
threshold is a choice between the amount of noise reduction
and preservation of fringe lines. The thresholded image
is shown in Fig. 4d. This process could also have been
done automatically by using the histogram. This requires
a bimodal histogram, which means that the grey-level
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data 281
distribution has two peaks, representing the noise
and lines respectively. The threshold could then have
been chosen at the frequency minima between the
peaks; see Gonzales & Woods (1993). However, in the
histogram of this image there is a smooth transition
between the noise pixels with the greatest values and
line pixels with the lowest values, and hence the histogram
is not bimodal.
The thresholded image is then vectorized with an existing
algorithm in ARC. Points that are closer than a certain
distance are connected with vectors. The resulting data
have to be edited to some extent. This is similar to the
edge-segment linking performed by Lin et al. (1994).
During the automatic editing, short individual line
segments are removed and small gaps in longer lines are
snapped. Some of the holes in the line segments are
longer than the smallest distance between two different
fringe lines, and cannot be snapped. It is therefore not
possible to close all the gaps automatically and some
manual editing has to be done. The vectorized fringe lines
are shown in Fig. 4e.
It is important for the calculations to determine the right
boundary for the glacier. The boundary is digitized manually
and used in all the steps following the median filtering to
limit the amount of calculations that have to be done. It also
serves as a reference line of zero velocity in the unwrapping.
It is best to use one of the amplitude images to determine the
glacier boundary; the DEM and the DInSAR scene are more
difficult to apply.
When the lines are edited, the topology for the polygons
that mark the transitions between the fringes can be created.
The polygons are numbered according to their relative
position. The outermost polygon is given the value 0,
the next 1, etc. This information is used in the unwrapping
of the InSAR image. Where half a wavelength (2.83 cm)
is added to the median-filtered InSAR-scene values inside
fringe one, one wavelength is added inside fringe 2, etc. The
median-filtered InSAR-scene is used because it contains
less noise and is believed to be closer to the real phase
differences than the original DInSAR scene. In other studies
it has been common to use the unfiltered interferogram
(Lin et al. 1994). The generation of topology could ideally
have been done automatically, but since some interpretation
is needed, it is done manually. The unwrapped scene,
showing displacements in centimetres towards the satellite,
is shown in Fig. 4f.
The fringe lines found by edge-detecting techniques will
never hit all the fringe transitions exactly. Some pixels that
should have been on the outside are therefore caught on the
inside, and vice versa. Pixels that end up on the wrong side
will differ greatly in pixel value from their neighbours after
the unwrapping. They will have pixel values that differ by
about half a wavelength from their neighbours. It is therefore
possible to remove these wrongly classified pixels by
statistical comparison of a pixel and its neighbours. For
reasons of simplicity, a median filter is used for this task as
well. The median filtering will alter all the pixels in the
image. However, since the variation of the data showed a
rather continuous trend, the altering will be relatively small
for the correctly classified pixels.
Since the scenes are already height differentiated, only one
part of the expression in Eq. 2 is left, namely the contribution
of glacier motion to the phase difference. It could then be
written as (Kwok & Fahnestock 1996):
Dfmot =
4p
4p
Dr = ~
vDT~
r
l
l
(Eq: 3)
where ~
v is the surface ground velocity in the direction of the
radar beam, DT the time between the acquisitions (temporal
baseline) and ~
r the unit vector in direction of range. The
unwrapped values, explained in the proceeding section,
represent the displacement towards the satellite. A displacement calculated towards a satellite is of limited use from a
glaciological point of view. It is therefore common to assume
surface parallel flow and decompose the motion into the
direction of surface slope.
Since the Norwegian Polar Institute had DEMs available,
the velocities were decomposed in the flow direction of the
glacier. Terrain slope is calculated together with aspect (i.e.
direction of steepest surface slope). This is done automatically with a gradient filter. Before the terrain parameters
could be calculated, substantial smoothing of the DEM had
to be done, since the glacier flow direction is governed by
large-scale changes in slope and aspect, and not small
undulations in surface topography. Kamb & Echelmeyer
(1986) found that, for modelling glacier velocities using
stress-gradient coupling, the filters had to be 4 to 10 times the
ice thickness. Using the 273 m depth found by Bamber
(1987) in the area between Kronebreen and Isachsenfonna,
the work by Kamb & Echelmeyer (1986) implies the
application of smoothing filters covering up to 2.7 km.
The radar look angle was also calculated. It varies between
20 and 26 from near-range to far-range.
When these three parameters, the slope of the glacier, the
aspect angle between glacier slope and radar range direction,
and the radar look angle, have been calculated, together
with the unwrapped interferogram, it is possible to calculate
the velocity, Vglac, in the flow direction of the glacier
(K. Eldhuset, personal communication 1998):
Vglac =
Uw
cos a cos f sin y þ cos y sin a
(Eq: 4)
Uw is the displacement values in the unwrapped scene, a is
the slope of the glacier, f is the aspect angle between glacier
slope and radar range direction, and y is the radar look angle.
Generally, there should also have been a DT (temporal
baseline) term in the denominator of Eq. 4, but since it is
tandem data and velocities are calculated in metres per day
(m d1), this term is omitted. The geometry for decomposition of glacier movement is shown in Fig. 5. Since the glacier
velocity in this study is defined as positive in the downward
direction of the steepest slope, the denominator of Eq. 4 has a
plus sign differing from the minus sign found in corresponding equations used in other studies, defining the velocity
to be positive upwards (Kwok & Fahnestock 1996). This
difference is only due to the definition of velocity direction
and has no implications for the result. All the parameters are
calculated for every pixel at the glacier surface, resulting in a
complete velocity field for the whole glacier.
282 B. Wangensteen, D.J. Weydahl & J.O. Hagen
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Discussion
Satellite
z
Since the goal of this study was to develop a method, a
detailed glaciological discussion of the results is beyond the
scope of this article. However, the results can be used to
assess the accuracy and the limitations of the method in
glaciological applications.
x
Vglac
r
Accuracy
azimuth
y
Vglac
x
range
~glac is surface
Fig. 5. The geometry used for decomposing glacier velocities. V
parallel, hence, a is the surface slope, and f is defined as the aspect angle.
Results
The described method has been used to calculate the surface
velocities along the surface-flow direction for three glaciers
at Svalbard. Data are presented in Table 2. The velocity field
based on the April 1996 scene of Isachsenfonna is shown
in Fig. 6a. The maximum velocity is 0.42 m d1 in the
images acquired in January 1996, having an average velocity
for the whole glacier surface of 0.23 m d1. For the images
acquired in April 1996, the maximum velocity is 0.42 m d1
and the average velocity is 0.18 m d1.
For Nordbreen, the maximum velocities vary between
0.33 and 0.36 m d1 and the average velocities vary between
0.10 and 0.17 m d1. The maximum velocities are found
in the areas of the greatest slope. The velocity field derived
from the InSAR-scene of Nordbreen acquired in April 1996
is shown in Fig. 6b.
The velocity field derived from the InSAR-scene of
Akademikerbreen is shown in Fig. 6c. The maximum
velocity is 0.41 m d1 and the average velocity is
0.07 m d1. There is a significant increase in the velocity
in the narrow passage between the two nunataks in the centre
of the scene. In this passage, the surface slope is also greater
than for the rest of the glacier.
The accuracy of the phase measurements with satellite
SAR interferometry is believed to be 1.5 mm for vertical
movements and 4 mm for horizontal movements (Goldstein
et al. 1993). These accuracy values are in the direction of the
radar beam (range) and will be larger for decomposed
velocities. The errors due to unwrapping are considered
negligible since local errors do not propagate across the
fringe lines, and the filtering has removed most of the local
noise. Errors in the DEM caused by interpolation and
co-registration will affect the resulting interferogram.
Eldhuset et al. (2003) found RMS differences between the
DEM and the InSAR-derived DEMs, using the same tandem
scenes as in this study, of 7 m to 20 m. This RMS difference
is an expression of errors in the DEM, errors in the
InSAR-derived DEM, and also surface elevation changes
between the acquisitions of the two different DEMs. The
average relative error in the DEM used for height differentiation is believed to be between 5 and 10 m, based on air
photo scale and general knowledge on the map compilation
and DEM interpolation procedure (T. Eiken, personal
communication 2004). A relative error of 10 m in the DEM
will give different errors for the end result for each glacier
depending on baseline, terrain slope and glacier-satellite
geometry. The errors for the different glaciers are in
the range 1–10% of the calculated average velocities
(Table 2).
Variation in water vapour in the atmosphere between
the two acquisitions can also have an impact on the phase
measurements, and hence also on the deformation measurements. Zebker et al. (1997) found that temporal and spatial
variation in air humidity of 20% could cause errors in the
deformation calculation of up to 10 cm. If such a variation
of humidity is the case on Svalbard the interferogram will
probably be unusable due to decorrelation. Several of the
DInSAR-scenes available were unusable since a small
change in humidity also can cause big differences in the
backscatter properties of snow. Due to Svalbard’s geographical position, large and fast fluctuations in temperature
Table 2. Data from the velocity calculations showing maximum (Vmax) and average (Vaver) velocities, the calculated average terrain slope and average aspect
(i.e. angle between range and the direction of steepest slope), and error of calculated glacier velocities caused by relative errors in the DEM (Etopo).
Glacier
Isachsenfonna
Isachsenfonna
Akademikerbreen
Nordbreen
Nordbreen
Nordbreen
Location
79 N 13 E
79 N 13 E
79 N 19 E
80 N 16 E
80 N 16 E
80 N 16 E
Date (ERS1/ERS2)
Vmax (m/d)
Vaver (m/d)
Slope (deg.)
Aspect (deg.)
Etopo (%)
10/11.10.96
05/06.04.96
20/21.10.95
27/28.09.95
05/06.04.96
10/11.05.96
0.42
0.42
0.41
0.36
0.36
0.33
0.23
0.18
0.07
0.17
0.13
0.10
0.46
0.45
1.19
1.02
1.14
1.07
128.0
128.0
176.0
43.7
42.4
17.0
2.3
0.9
3.1
9.8
1.2
6.4
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data 283
Fig. 6. The resulting velocity fields of Isachsenfonna 5–6 April 1996 (a), Nordbreen 5–6 April 1996 (b), and Akademikerbreen 27–28 October 1995 (c).
and weather conditions occur. This can be the reason for the
high percentage of decorrelated DInSAR-scenes.
The penetration depth of radar signals varies a lot. The
ERS-1 and ERS-2 signal with a frequency of 5.3 GHz could
penetrate 10 m of ice and some tens of metres of dry snow
(Ulaby et al. 1982). Hoen & Zebker (2000) reported
penetration depths of 12 m to 35 m from ERS data from
the Greenland Ice Sheet. It could therefore be stated that, in
fact, it is not surface velocity that is measured. However,
in accordance with glaciological theory, most of the vertical
velocity change in a glacier takes place close to the bed.
Bamber (1987) found an ice thickness of 273 m in the area
between Kronebreen and Isachsenfonna. This means that
there is practically no difference in the surface velocity and
the velocity found some tens of metres down in the glacier
(Paterson 1994, 251). If the penetration causes the signal to
be reflected from a level lower than the level of the external
DEM, this will create errors in the DInSAR. However, most
Svalbard glaciers will experience epochs of melting even
during the winter. Ice layers and ice bodies within the snow
will therefore cause most of the reflection to happen near the
surface.
The estimation of the terrain parameters also introduces
inaccuracies, but because of the substantial smoothing of the
DEM it is not influenced much by the relative errors in the
DEM described previously. However, surface slope estimation accuracy has an impact on the end result but not as
drastically as the direction of slope. The average surface
slopes in the three InSAR-scenes vary between 0.45 and
1.19 (Table 2) and for these small angles an error will not
cause much harm according to Eq. 4. Yet when the direction
of slopes approaches the azimuth direction the accuracy
drastically decreases. This is due to the fact that a DInSARscene only measures the displacement component in range.
Fig. 7 shows how the calculated velocity will vary with an
increased angle between the direction of slope and range.
An error of 10 if the direction of slope is 0 will cause small
errors (1.54%) for the calculated velocities. On the other
hand, an error of 10 for a glacier moving at an angle of
50 with range will be much higher (up to 46.2%). These
284 B. Wangensteen, D.J. Weydahl & J.O. Hagen
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
6000
Glacier velocity
5000
Comparing the results from Lefauconnier et al. (2001), it is
evident that the velocities differ somewhat from the
velocities obtained for Isachsenfonna in this study. They
found a nearly constant velocity for the whole glacier,
ranging from 0.20 to 0.22 m d1, while the results presented
here show a change along the centre line from c.0.20 m d1
in the higher parts (lower left in Fig. 6a), via maximum
velocities of over 0.40 m d1 around the middle, and down
to 0.15–0.20 m d1 in the lower parts (upper right in
Fig. 6a). Calculations further down the glacier are omitted
in this study since the flow direction is turning more
southwards and more or less parallel to the azimuth direction,
and hence velocity measurements will be heavily influenced
by errors or even impossible. On the lower left side of the
glacier there is an expected decrease in velocity. This is
an area in between the ice flows of Holtedahlsfonna and
Isachsenfonna where one would expect stagnant ice masses.
Lefauconnier et al. (2001) also report on GPS- and InSARmeasured velocities close to this area of 0.10 and 0.8 m d1
respectively.
The difference in measured velocities at the mid to upper
part of Isachsenfonna could be caused by the fact that
the glacier is at a relatively small angle with the azimuth
direction in the InSAR scene used in Lefauconnier et al.
(2001). They also state that they had some problems with
fringe counting and that only the velocities measured in the
central basin (i.e. where Isachsenfonna and Holtedahlsfonna
merge) should be considered valid. Looking at the geometry
of the glacier, a velocity increase in the narrow part of the
glacier and a decrease further downstream could be expected
where the glacier is widening, in common with those
presented here. More ground measured velocity data have
to be gathered to confirm this.
Generally, all the velocity maps in this study show
velocity patterns that conform to glaciological theory;
velocity increases with surface slope and decreases with
increasing width. Thus the glaciers show extending and
compressive flow due to changes in surface slope and width
as expected. These factors overrun the climatic ones and
therefore there is no good correlation between the position
of the equilibrium line and the maximum velocities. The
greatest change in velocity is found along the edges of the
glacier, since the edges exert a drag on the glacier.
For Nordbreen, two late winter scenes and one autumn
scene were chosen. The highest velocities are found in
the September 1995 scene, which also has a higher mean
velocity than the other scenes. This could indicate higher
velocities during summer due to more meltwater. Since the
scene was acquired in late September in the Arctic, this is
probably not the case. Hence, there has to be an alternative
explanation for the increase. The autumn scene was also
acquired the year before the spring scenes, so the explanation
could equally be one other than seasonal variation. The
uncertainty in determining the glacial border for Nordbreen
can also have implications for the result. To be able to
determine a seasonal variation, a scene from July or August
is needed together with a better determination of the glacier
border.
%Δv
4000
3000
2000
1000
25
50
75
100
125
150
175
100
%Δv
80
60
40
20
10
20
30
40
50
60
Fig. 7. Relative change of decomposed glacier velocity as a function of the
angle between direction of slope and range (i.e. aspect). The upper graph
shows relative change in decomposed velocity for all angles and a great
relative change close to 90 (i.e. azimuth direction). The lower graph shows
the same property for aspect angles between 0 and 60 . The aspect angles for
the glaciers investigated in this study are between 4 and 52 .
estimates are also based on Eq. 4. Due to this, Akademikerbreen will have the most accurate results, since the average
angle between range and the direction of slope is rather small,
as shown in Table 2. For the same reason, the accuracy of the
estimated velocity for Nordbreen and Isachsenfonna will be
lower, and calculation for the lower parts of Isachsenfonna is
omitted completely. The greatest mean aspect angle is 52 .
Since it is difficult to establish the accuracy of the slope
direction calculation, it is also difficult to say anything more
precisely about the effect on the calculated velocities. If the
slope direction angle of 52 has a 5% error, the error after
decomposition is 6.3%, while a 10% error will cause a 13.7%
error in the end result. Rignot et al. (1995) determined the
error in the direction of slope in their study to be +4 . If it is
assumed that the accuracy of slope calculation in this study is
0.5 , the look angle 0.2 , and direction of slope 5%, then the
total error from decomposition will be just under 7%. The
error for glaciers moving in the range direction will be smaller,
close to 3%, since the error from the direction of slope
calculation will be much smaller. In addition, there will still be
errors from the aforementioned phase measurements, of
1.5 mm for vertical and 4 mm for horizontal movement, and
also errors from differentiating the InSAR scenes. Rignot et al.
(1995) found InSAR-calculated velocities to be within 6% of
ground measured ones.
NORSK GEOGRAFISK TIDSSKRIFT 59 (2005)
Mapping glacier velocities on Svalbard using ERS tandem DInSAR data 285
Conclusions
The results from this study show that glacier velocities
on Svalbard can be calculated with DEMs and heightdifferentiated tandem InSAR data by the use of modules in
ESRI’s Arc GIS software. A semi-automatic algorithm is
developed in order to calculate glacier velocities in the flow
direction, resulting in maps of the velocity fields for the
glaciers. Velocity fields for the three glaciers Isachsenfonna,
Nordbreen and Akademikerbreen are presented, with maximum values of 0.42 m d1, 0.36 m d1 and 0.41 m d1
respectively. The accuracy is greatest when the angle
between direction of slope and range is small. The results
seem to be reliable and in agreement with other ground and
satellite-borne observations.
The results also show that the ERS-1/ERS-2 SAR tandem
archive from 1995–1996 can reliably be used to establish
velocity fields for most of the larger glaciers on Svalbard. Such
a database showing the state of the glaciers in the mid-1990s
will work as a unique reference when making new field
measurements or remote sensing acquisitions in the future.
Acknowledgements. – The authors thank ESA for supporting the SAR raw
data that originally was delivered as part of the ESA Announcement of
Opportunity (AO) Tandem project number AOT.N301 that was carried out at
the Norwegian Defence Research Establishment (FFI). We also thank the
Norwegian Polar Institute for providing the DEM. The authors also wish
thank the reviewers for their valuable comments.
Manuscript submitted 23 September 2004; accepted 13 January 2005
References
Bamber, J.L. 1987. Radio Echo Soundings of Svalbard Glaciers. PhD thesis,
Jesus College, Cambridge.
Eldhuset, K., Aanvik, F., Aksnes, K., Amlien, J., Andersen, P.H., Hauge, S.,
Isaksson, E., Wahl, T. & Weydahl, D.J. 1996. First Results from InSAR
Processing on Svalbard. ESA SP-406: Conference proceedings of Fringe
’96 (Zürich, Switzerland 30 September – 2 October). European Space
Agency, Munich.
Eldhuset, K., Andersen, P.H., Hauge, S., Isaksson, E. & Weydahl, D.J. 2003.
ERS tandem InSAR processing for DEM generation, glacier motion
estimation and coherence analysis on Svalbard. International Journal of
Remote Sensing 24:7, 1415–1437.
Gens, R. & van Genderen, J.L. 1996. SAR interferometry – issues,
techniques, applications. International Journal of Remote Sensing 17:10,
1803–1835.
Goldstein, R.M., Engelhardt, H., Kamb, B. & Frolich, R.M. 1993. Satellite
radar interferometry for monitoring ice sheet motion: Application to an
Antarctic ice stream. Science 262:5139, 1525–1530.
Gonzalez, R.C. & Woods, R.E. 1993. Digital Image Processing. AddisonWesley, Reading, MA.
Hagen, J.O., Liestøl, O., Roland, E. & Jørgensen, T. 1993. Glacier Atlas of
Svalbard and Jan Mayen. Norsk Polarinstitutt Meddelelser 129. Norsk
Polarinstitutt, Oslo.
Hamilton, G.S. & Dowdeswell, J.A. 1996. Controls of glacier surging in
Svalbard. Journal of Glaciology 42:140, 157–168.
Hoen, E.W. & Zebker, H.A. 2000. Penetration depths inferred from
interferometric volume decorrelation observed over the Greenland ice
sheet. IEEE Transactions on Geoscience and Remote Sensing 38:6,
2571–2583.
Jiskoot, H., Murray, T. & Boyle, P. 2000. Controls on the
distribution of surge-type glaciers in Svalbard. Journal of Glaciology
46:154, 412–422.
Joughin, I.R., Winebrenner, D.P. & Fahnstock, M.A. 1995. Observations of
ice-sheet motion in Greenland using satellite radar interferometry.
Geophysical Research Letters 22:5, 571–574.
Kääb, A. & Funk, M. 1999. Modelling mass balance using photogrammetric
and geophysical data: A pilot study at Griesgletcher, Swiss Alps. Journal
of Glaciology 45:151, 575–583.
Kamb, B. & Echelmeyer, K.A. 1986. Stress-gradient coupling in glacier flow:
I. Longitudinal averaging of the influence of ice thickness and surface
slope. Journal of Glaciology 32:111, 267–281.
König, M., Winther, J.G. & Isaksson, E. 2001. Measuring snow and glacier
ice properties from satellite. Reviews of Geophysics 39:1, 1–27.
Kwok, R. & Fahnestock, M.A. 1996. Ice sheet motion and topography from
radar interferometry. IEEE Transactions on Geoscience and Remote
Sensing 34:1, 189–200.
Lefauconnier, B. & Hagen, J.O. 1991. Surging and Calving Glaciers
in Eastern Svalbard. Norsk Polarinstitutt Meddelelser 116. Norsk
Polarinstitutt, Oslo.
Lefauconnier, B., Hagen, J.O. & Rudant, J.P. 1994. Flow speed and calving
rate of Kongsbreen glacier, Svalbard, using SPOT images. Polar Research
13:1, 59–65.
Lefauconnier, B., Massonnet, D. & Anker, G. 2001. Determination of ice
flow velocity on Svalbard from ERS-1 interferometric observations.
Watanabe, O. & Yamanouchi, T. (eds.) Proceedings of Second International Symposium on Environmental Research in the Arctic and Fifth
Ny-Ålesund Seminar 23–25 February 2000, NIPR, Tokyo. Memoirs of
National Institute of Polar Research, Special Issue 54, 279–290. National
Institute of Polar Research, Tokyo.
Liestøl, O. 1969. Glacier surges in West Spitsbergen. Canadian Journal of
Earth Science 6, 895–897.
Liestøl, O. 1977. Pingos, springs, and permafrost in Spitsbergen. Norsk
Polarinstitutt Årbok 1975, 7–29.
Liestøl, O. 1988. The glaciers in the Kongsfjorden area, Spitsbergen. Norsk
Geografisk Tidsskrift 42, 231–238.
Lin, Q., Vesecky, J.F. & Zebker, A. 1994. Phase unwrapping through fringe
line detection in synthetic aperture radar interferometry. Applied Optics
33:2, 201–208.
Massonnet, D. & Feigl, K.L. 1998. Radar interferometry and its
application to changes in the earth’s surface. Reviews of Geophysics
36:4, 441–500.
Michel, R. & Rignot, E. 1999. Flow of Glaciar Moreno, Argentina, from
repeat-pass Shuttle Imaging Radar images: Comparison of the phase
correlation method with radar interferometry. Journal of Glaciology
45:149, 93–100.
Mohr, J.J., Reeh, N. & Madsen, S.N. 1998. Three-dimensional glacial
flow and surface elevation measured with radar interferometry. Nature
391:6664, 273–276.
Paterson, W.S.B. 1994. The Physics of Glaciers. Pergamon/Elsevier Science,
Oxford.
Rignot, E., Jezek, K.C. & Sohn, H.G. 1995. Ice flow dynamics of the
Greenland Ice Sheet from SAR Interferometry. Geophysical Research
Letters 22:5, 575–578.
Rott, H. & Siegel, A. 1996. Glaciological Studies in the Alps and in
Antarctica Using ERS Interferometric SAR. ESA SP-406: Conference
proceedings of Fringe ’96 (Zürich, Switzerland 30 September–2 October).
European Space Agency, Munich.
Ulaby, F.T., Moore, R.K. & Fung, A.K. 1982. Microwave Remote
Sensing: Active and Passive Radar Remote Sensing and Surface
Scattering and Emission Theory. Vol. 2. Addison-Wesley, Reading,
MA.
Unwin, B. & Wingham, D. 1997. Topography and dynamics of Austfonna,
Nordaustlandet, Svalbard, from SAR interferometry. Annals of Glaciology
24, 403–408.
Voigt, U. 1965. Die Bewegung der Gletscherzunge der Kongsvegen,
Kingsbay, Vestspitsbergen. Petermanns Geographische Mitteilungen
109:1, 1–8.
Weydahl, D.J. 2001. Analysis of ERS tandem SAR coherence from glaciers,
valleys, and fjord ice on Svalbard. IEEE Transactions on Geoscience
and Remote Sensing 39:9, 2029–2039.
Zebker, H.A., Rosen, P.A. & Hensley, S. 1997. Atmospheric effects
in interferometric synthetic aperture radar surface deformation and
topographic maps. Journal of Geophysical Research, B, Solid Earth and
Planets 102:B4, 7547–7563.