See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249552014

Aerial photography and digital photogrammetry for landslide monitoring

Article  in  Geological Society London Special Publications · January 2007

DOI: 10.1144/SP283.5

CITATIONS READS

13 1,678

4 authors:

Jan Walstra Jim H. Chandler

Royal Belgian Institute of Natural Sciences Formely at Loughborough University
31 PUBLICATIONS   221 CITATIONS    127 PUBLICATIONS   3,875 CITATIONS   

SEE PROFILE SEE PROFILE

Neil Dixon Tom Dijkstra

Loughborough University Loughborough University
135 PUBLICATIONS   1,742 CITATIONS    74 PUBLICATIONS   969 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

GEPATAR - GEotechnical and Patrimonial Archives Toolbox for ARchitectural conservation in Belgium View project

Collapsible loess soils View project

All content following this page was uploaded by Jan Walstra on 03 February 2014.

The user has requested enhancement of the downloaded file.

 Aerial photography and digital photogrammetry
for landslide monitoring
J. WALSTRA, J. H. CHANDLER, N. DIXON & T. A. DIJKSTRA
Department of Civil & Building Engineering, Loughborough University,
Loughborough LE11 3TU, UK (e-mail: [email protected])

Abstract: A review is given of the techniques that are available to extract relevant information
from multi-temporal aerial photographs for use in the monitoring stage of landslide assessments.
It is shown that aerial photograph interpretation reveals qualitative information on surface
characteristics, which is helpful in detecting landslide features and inferring the mechanisms
involved. Photogrammetrically derived products can be used to quantify these processes, provid-
ing distinctive advantages. Comparison of digital elevation models (DEMs) from different times
provides detailed information on changes in surface topography, whereas orthophotos can be used
to measure horizontal displacements. The various factors influencing the quality of the products
are also identified. Examples from a case study on the Mam Tor landslide are used to illustrate
the benefits of the different approaches.

Aerial photographs are a generally accepted tool diagnostic surface features, such as morphology,
used in landslide studies. They not only provide vegetation cover, soil moisture and drainage
a metric model from which quantitative measure- pattern. Furthermore, recent photographs can be
ments can be obtained, but also give a qualita- compared with historical imagery to assess land-
tive description of the Earth surface. These two slide conditions over different periods of time and
capabilities are irrefutably related to each other, allow the progressive development to be examined.
as ‘one must know what one is measuring’ (Lo Characteristics of mass movements that can be
1976). monitored by sequential photographs are, for
The application of aerial photographs to land- example, the areal extent of the landslide body,
slide investigation provides a number of distinct regression rate of the head scar, displacement velo-
advantages. Reconnaissance of the study area can city, surface topography, succession of vegetation
greatly benefit from the 3D representation that and soil moisture conditions. Accurate quantifi-
is provided by stereoscopic viewing, thereby cation of change requires the application of rigorous
showing relationships between the various land- photogrammetric techniques (Chandler 1989).
scape elements more obviously than from a Finally, aerial photography can be helpful in
ground perspective. Furthermore, photographically hazard mapping. The purpose of landslide hazard
based derivatives provide a suitable base on which mapping is to analyse the susceptibility of the
boundaries can be delineated accurately. In terrain to slope movements. Aerial photographs
addition, photographs support the efficient planning can be used to delimit terrain units and map the
of field investigations and sampling schemes, controlling factors affecting slope stability.
without the need for visiting the site physically, The aim of this paper is to give an overview of
which is especially useful in remote and inaccess- the ways in which aerial photographs and digital
ible areas (Crozier 1984, Van Zuidam 1985). A photogrammetric techniques can be used in the
final and important advantage is the quantitative monitoring stage of landslide assessments. Particu-
topographic information contained, which can lar attention will be paid to the quality of data
be unlocked by appropriate photogrammetric derived from aerial photographs of differing type,
techniques. However, quantitative use of aerial when using the different techniques available. The
photographs create some difficulties, such as the various approaches can be roughly divided into
requirement of experienced analysts and appropri- three categories: those based on simple aerial
ate equipment, combined with sufficient knowledge photograph interpretations (APIs), those involving
of the site under investigation (Lo 1976). the extraction of digital elevation models (DEMs),
Aerial photographs can be used in various stages and those based on the creation of orthophotos.
of landslide investigations (Mantovani et al. 1996), The underlying techniques will be described and
and have been extensively used in the detection illustrated with some results that were obtained
and classification of landslides. When properly from a case study focusing on the Mam Tor land-
interpreted they allow the identification of slide (Derbyshire, UK).

From: TEEUW , R. M. (ed.) Mapping Hazardous Terrain using Remote Sensing. Geological Society,
London, Special Publications, 283, 53– 63.
DOI: 10.1144/SP283.5 0305-8719/07/$15.00 # The Geological Society 2007.
54 J. WALSTRA ET AL.

The study area: Mam Tor The unstable transition zone, overlying the steepest
part of the basal shear, is the most active part,
The landslide of Mam Tor is situated on the eastern moving on average 0.35 m a21 over the last
flank of this 517 m high hill, at the head of the Hope century (Rutter et al. 2003). There is evidence
Valley, Derbyshire, UK [SK135835]. The former that the movements are not continuous but accele-
main road between Sheffield and Manchester rate during wet winters, when rainfall exceeds
(A625) was constructed across the slide, but certain limits; that is, more than 250 mm rain in a
abandoned in 1979 as a consequence of continuous single month and over 750 mm in the preceding 6
damage caused by the moving ground mass (Fig. 1). months (Waltham & Dixon 2000).
The slope consists of predominantly sandstone There are several information sources available
sequences (Mam Tor Beds) overlying predomi- that quantify displacements that have taken place
nantly shale units (Edale Shales). The layers dip over the last century. Notes about regular dis-
slightly inwards of the slope. From scarp to toe, turbance and repairs of the road, from 1907 until
the landslide measures c. 1000 m, and elevation the final closure in 1979, are kept by Derbyshire
varies from 510 to 230 m. The mean slope of the and stability analysis was carried out (Skempton
slipped mass is 128 and the maximum thickness et al. 1989). Since closure of the road, temporary
30–40 m (Skempton et al. 1989). monitoring schemes were set up by Sheffield
The initial rotational failure has been dated back University (1981–1983; Al-Dabbagh & Cripps
to 3600 BP (Skempton et al. 1989). While advan- 1987), Nottingham Trent University (1990–1998;
cing downslope the mass broke into a complex of Waltham & Dixon 2000) and Manchester Univer-
blocks and slices. Disintegration of the front slices sity (since 1996; Rutter et al. 2003).
created a debris mass, which slid further down.

Aerial photographs
A conventional photo search for aerial photography
of Mam Tor revealed that there are numerous image
epochs available, both oblique and vertical, from
1948 until the present. Vertical imagery from
eight epochs was acquired and processed
(Table 1). The images are of varying quality and
scales, and can be used in an assessment of the
potential of the various techniques applied to a
range of commonly available material.
Photogrammetric data processing was achieved
using the IMAGINE OrthoBASE Pro 8.6
software package (ERDAS LLC 1991–2002).
During photogrammetric processing the relation-
ship between photo co-ordinates and the Ordnance
Fig. 1. Damaged road section at Mam Tor. Survey national grid co-ordinate system was

Table 1. Characteristics of the acquired image epochs of Mam Tor

Date Source Scale Scan resolution Ground resolution Media

(mm) (m)

1953 NMR* 1/10 700 42 0.45 Scanned contact prints

1971 NMR 1/6 400 42 0.27 Scanned contact prints
1973 CUCAP† 1/4 300 15 0.065 Scanned diapositives
1973 CUCAP Oblique 15 – Scanned diapositives
1984 ADAS‡ 1/27 200 15 0.41 Scanned diapositives
1990 CUCAP 1/12 000 15 0.18 Scanned diapositives
1995 CUCAP 1/16 400 15 0.25 Scanned colour negatives
1999 Infoterra 1/12 200 21 0.26 Scanned colour negatives

*National Monuments Record.

†Cambridge University Collection of Air Photos.
‡Agricultural Development and Advisory Service.
 PHOTOGRAPHY FOR LANDSLIDE MONITORING 55

established. An independent module performing lower flying height increases the base/distance
a self-calibrating bundle adjustment was used for ratio. However, this may create problems of ‘dead
estimating the camera’s interior parameters ground’ on far side of hills and in narrow valleys.
(camera constants), if the original calibration certi- Oblique photographs sometimes can provide a
ficate was unavailable (as described by Chandler & more unobstructed view of steep slopes and cliffs
Clark 1992). Ground control was collected by (Rib & Liang 1978), and give a more familiar
means of a differential global positioning system perspective for the less experienced interpreter
(GPS) survey. (Chandler 1989).
The most suitable scale is inevitably a com-
promise. Large-scale photographs provide a high
Aerial photograph interpretation (API) level of detail, but may require many frames to
cover the study area. Small-scale photography pro-
Photo-interpretation involves the systematic exami- vides less detail, but allows a better interpretation
nation of photographic images for the purpose of of the overall context. Quoted optimum scales for
identifying objects and judging their significance site studies are in the range between 1/5000 and
(Colwell 1960). Although aerial photographs can 1/15 000 (Norman et al. 1975; Mantovani et al.
be interpreted with a specific theme in mind, 1996; Soeters & Van Westen 1996).
interpretation relies on using the same basic charac- The time of the day when photographs are taken
teristics of the surface: tone, texture, pattern, shape, determines the length of shadows. In general,
context and scale, which were created by reflection photographs taken when the sun is high and
of natural electromagnetic light energy from the shadows on the hillsides and slopes are minimal
objects that make up the scene and their arrange- are best for interpretation. However, in areas of
ment. The use of these qualitative attributes is low topography, the relief will be enhanced by
very much a matter of experience and personal long shadows. The time of the year influences the
bias (Drury 1987). effects of soil moisture and vegetation (Norman
The quality of an API is affected by several et al. 1975; Soeters & Van Westen 1996). The
factors, which can be separated into four main cate- quality of photographs depends on the various
gories: photographic parameters, natural factors, processes the images go through. Norman et al.
equipment and analysis techniques, and the qualifi- (1975) used the following criteria for assessing
cation of the interpreter (Rib & Liang 1978). photo quality: sharpness, over- or under-exposure,
cloud cover, shadow and print quality.
Photographic parameters In spite of recent developments in the field of
airborne digital sensors (Eckhardt et al. 2000), the
The effects of the different photographic parameters most common way of obtaining digital imagery is
on landslip detection have been described by by scanning the original film. Photogrammetric
Norman et al. (1975) and Soeters & Van Westen scanners have a high geometric resolution, but
1996). Natural colour and panchromatic (black- radiometric performance may be rather poor
and-white) films are the most widely available (Baltsavias 1999). Modern software packages
film types. Colour film is especially valuable for allow digital images to be easily adjusted to the
outlining differences in soil conditions, drainage needs of the user, for example, zooming in on
and vegetation. Colour IR films are most suitable particular areas or enhancing the contrast.
for detecting landslides, mainly because of the
capability of identifying the presence of water and Natural factors
thus showing the vigour of vegetation cover.
Panchromatic films, on the other hand, provide a Photo-interpretation is also influenced by natural
better image resolution (Lo 1976) and are generally factors. Steep slopes, forest canopy and shadows
less expensive. Most historical imagery is of this may hide certain surface features. Optimal con-
form, although resolution tends to degrade with ditions for detecting anomalies in vegetation may
photo age, as a result of developments in photo- be expected in either the very early or very late
graphic emulsion that have subsequently occurred. stages of the growing season. Differences in drain-
Aerial photographs for mapping purposes are age conditions are most pronounced shortly after
typically vertical, with 60% overlap between suc- the start of the wet season or shortly after the snow-
cessive frames to provide stereoscopic coverage. melt period in spring (Soeters & Van Westen 1996).
Stereoscopic viewing is important, as landslide Weather conditions have an important influence on
features are most frequently recognized by their photo quality: clouds and snow cover may obscure
morphology. Vertical exaggeration, when viewing the ground surface, haze decreases contrast, and
stereoscopically, can be enhanced if a super wide solar angle influences shadowing (Rib & Liang
angle lens is used during photo acquisition. The 1978).
56 J. WALSTRA ET AL.

Qualification of interpreter these aspects can be depicted by coloured area

symbols, patterns and line symbols (Van Zuidam
The quality of API is also influenced by the 1985). A great variety in geomorphological mapping
capability of the human interpreter, particularly systems have been designed in the course of time.
experience in photo-interpretation, and knowledge Savigear 1965) developed a purely morphological
of the phenomena and processes being studied. legend, aiming to describe the form of slopes,
Various researchers have shown the large subjec- without reference to the origin. The ITC system
tive element in photo-interpretation by comparing (Van Zuidam 1985), designed for multipurpose
maps of the same landslide area created by different use at all scales, distinguishes the highest level on
interpreters (Van Westen 1993; Carrara et al. 1995). the basis of morphogenesis. The importance of
Identification of the exact positions of morpholo- geomorphological mapping has also been recog-
gical features can be difficult, especially delineation nized by engineering geologists, judging from the
of the boundaries. Moreover, different classes may proposed legend for engineering geological maps
be assigned to a specific feature as a result of differ- by the Geological Society Engineering Group
ent interpretation (Chandler 1989). Obviously, Working Party, which contains a large number
different mapping legends will lead to very differ- of symbols for geomorphological features
ent maps (Van Westen et al. 1999). (Anonymous 1972).
API is a valuable tool in geomorphological
Diagnostic features mapping, although field work remains necessary
for checking the photo-interpretation and mapping
The interpretation of landslides from aerial photo- of small features (Hayden 1986). Accurate defi-
graphs is mainly based on features indirectly nition and coding of geomorphological boundaries
related to slope movements, such as characteristic by rigorous photogrammetric techniques combines
morphology, anomalous vegetation and drainage the benefits of geomorphological interpretation
conditions, or disturbed infrastructure. Soeters & with positional relevance (Chandler & Brunsden
Van Westen (1996) have provided an extensive 1995). Photogrammetric measurements allow
overview of terrain features associated with quantitative comparison between photo-interpreted
landslides and their characterization on aerial maps from different periods (Chandler & Cooper
photographs. Based on these diagnostic features, 1989). Multi-temporal geomorphological maps
statements can be made on the type of movement, from sequential aerial photographs have been used
degree of activity and depth of movement to document the evolution of the Black Ven
(Mantovani et al. 1996). landslide, UK (Chandler & Brunsden 1995), and
the Tessina landslide, Italy (Van Westen &
Geomorphological mapping Getahun 2003). Kalaugher et al. (1987) used
oblique aerial photographs to identify geo-
A useful tool in presenting photo-interpreted infor- morphological processes on sea cliffs in East
mation is a geomorphological map. Geomorpho- Devon, UK.
logical maps are transmitters of information about Figure 2 displays a simple geomorphological
the form, origin, age and distribution of landforms map of the Mam Tor landslide, created by photo-
together with their formative processes, rock type interpretation of the 1990 images. The mapped
and surface materials (Brunsden et al. 1975). features were placed in the exact spatial context
They are not only a way of presenting data, but by using photogrammetry. Contour lines were
also the result of a method of research, revealing obtained from a DEM, extracted from the same
associations of landforms, which is essential for photographs.
understanding of both individual landforms and
landscapes (De Graaff et al. 1987). Parise (2003)
pointed out the importance of large-scale geomor- Digital elevation models (DEMs)
phological mapping, and especially its repetition
in time, in the study of active mass movements. A Digital photogrammetric techniques have the
combination of detailed multi-temporal mapping capability of automatically deriving very high-
of surface features, indirect indicators of defor- resolution DEMs from stereo photographs, pro-
mation and displacements may result in better viding a detailed representation of the surface
understanding of the landslide and its zonation in topography (Chandler 1999). Photogrammetry is
different elements, characterized by different based on the concept of collinearity, whereby a
styles of deformation. point on the object, centre of lens and resultant
Geomorphological maps can emphasize differ- image point lie on a single line in 3D space.
ent aspects of landforms: form, origin, age and Based on this principle, 3D co-ordinates represent-
relations. Dependent on the purpose of the map, ing the object can be extracted from a stereopair of
 PHOTOGRAPHY FOR LANDSLIDE MONITORING 57

Fig. 2. A geomorphological map, created through photo-interpretation of the 1990 images.

photographs, provided that the inner geometry site, and used to create a DEM. A significant recent
(interior orientation) and the position and orien- development is the automation of this process.
tation of the camera at the moment of exposure Automatic generation of DEMs from a stereomodel
(exterior orientation) are known. The exterior orien- comprises three tasks: image matching, surface
tation parameters of all frames in a block can be fitting and quality control (Schenk 1996). The
simultaneously estimated in a bundle block adjust- process of image matching involves the identifi-
ment, with the help of ground control points of cation of conjugate points in the overlap portion
which both ground and image co-ordinates are of the images. A commonly applied matching
known (Wolf & Dewitt 2000). strategy is area-based cross-correlation, in which
Once the relationship between the photographs small image patches are compared according
and ground surface has been established, to their grey-level distribution. Perfect matches
co-ordinates can be extracted from anywhere on the will never occur in reality because of noise, small
58 J. WALSTRA ET AL.

differences in illumination, and small geometric height precision of photogrammetric measurements

distortions. Because a regular gridded DEM is is also dependent on the geometry provided by
often required, surface fitting needs to be per- stereo-photographs. A strong convergence (high
formed. This procedure comprises the interpola- base/distance ratio), and large relief displacement,
tion of intermediate points, as the points obtained gives rise to highly precise object coordinates
by image matching do not represent the entire (Wolf & Dewitt 2000). According to Fryer et al.
surface. (1994), the best vertical precision that can be
The quality of a DEM is a function of the accu- expected using standard mapping configurations is
racy, reliability and precision of the photogram- about 1– 3 parts per 10 000 of flying height.
metric measurements and the block bundle Systematic errors are always inherent in the
adjustment itself (Butler et al. 1998). As defined stereo-model, arising from a variety of sources
by Cooper & Cross 1988), precision is related including lens distortion, atmospheric effects, film
to random errors, inherent in the measurement deformation, scan distortions, and inaccurate or
process. The bundle adjustment procedure is poorly distributed control points, or result from
capable of propagating stochastic properties errors during the image matching procedure
through the model, providing an estimation of (Chandler 1989; Buckley 2003). If camera cali-
precision. Reliability can be related to gross bration parameters are not available, which is
errors. Fortunately, gross errors are normally easy sometimes the case when using archival imagery,
to detect and eradicate because of their size. these can be estimated in a self-calibrating
Accuracy is related to the presence of undetected bundle-adjustment. However, accounting for all
systematic errors, which are more difficult to systematic effects is difficult, because many syste-
isolate and generally provide a limiting constraint matic errors cannot be modelled explicitly,
on the quality of the derived data. The mean and and there is usually high correlation between the
standard deviation of the discrepancies with modelling parameters (Granshaw 1980).
independent check points provide a measure of Control points should be evenly distributed over
DEM accuracy (Butler et al. 1998). the images to gain a strong geometry. Ideal
locations tie frames together and surround the
Controls on DEM quality volume of interest. A minimum of two planimetric
and three height points is needed to define a
As pointed out by Fryer et al. (1994) and Lane et al. datum, but more control points are desirable as
(2000), the ease with which terrain data may be redundancy provides appropriate checks (Wolf &
generated using digital photogrammetric techniques Dewitt 2000). Automated image matching is
has focused attention more on analysis and inter- affected by surface texture and geometric distortion
pretation of the acquired results than on issues of caused by different viewing angle. These controls
data quality. In addition to the conventional controls upon automated generation of elevation data are
on photogrammetry, the automated algorithms in of special relevance to complex terrain surfaces
digital processing have important influences on (Lane et al. 2000). If there is insufficient texture,
the quality of results. the software is unable to match two points success-
The precision that can be achieved by photo- fully and an interpolated estimate may be created.
grammetric measurements is mainly dependent on Surface roughness has a positive effect on texture,
the quality of the source data (i.e. the aerial photo- and consequently on matching. However, this
graphs). Photographic resolution is a function of effect may be countered by the increasing differ-
the optical quality of an image, and influenced by ences in the viewing of areas, which thus reduce
the resolving power of the film and camera lens, the level of correlation between the images. In
image motion during exposure, atmospheric addition, interpolation will be least effective in
conditions and the conditions of film processing. areas of great roughness. DEM collection para-
The effects of scale and resolution can be combined meters can be optimized, but these control indi-
in terms of ground resolution distance, which vidual matches rather than affecting the resulting
determines the level of detail that is visible on the surface accuracy (Lane et al. 2000). Surface
photographs (Lillesand & Kiefer 1994). When quality is also affected by its point density. An
using digital images, scan resolution and the increase in grid spacing will smooth the topogra-
quality of the scanner (geometric and radiometric) phy; the minimum grid spacing is, however,
are important controls. To preserve an original bound by the object space pixel dimension (Lane
film resolution of 30 –60 lines per mm, a scanned et al. 2000).
pixel size of 6–12 mm would be needed. Gross errors are genuine mistakes or blunders
For many practical applications, such as DEM that arise during photogrammetric measurement
generation, good results can be achieved with (Cooper & Cross 1988). They can be detected by
25–30 mm resolution (Baltsavias 1999). The increasing the redundancy of measurements
 PHOTOGRAPHY FOR LANDSLIDE MONITORING 59

a DEM of one epoch from an earlier DEM

creates a grid surface representing the change of
form over the period. This surface of change,
or ‘DEM-of-difference’, quantifies the effects
of geomorphological processes. Areas experiencing
removal of material will be indicated by
depressions, whereas those receiving material
are indicated by peaks. Caution should be
taken, as areas exhibiting no change are not necess-
arily inactive regions; they can represent areas
where input of material has equalled output
during the time interval (Chandler & Brunsden
Fig. 3. A 3D view of Mam Tor, created by draping an 1995).
orthophoto over a DEM, obtained from the 1990 images. A photogrammetric method was developed in
the late 1980s that was able to derive quantitative
spatial information from historical aerial photo-
(Hottier 1976), which gives rise to datasets that are graphs (Chandler & Cooper 1989; Chandler &
‘internally reliable’ (Cooper & Cross 1988). Gross Clark 1992). This offered potential to unlock the
error sources that commonly affect the determi- photographic archive for obtaining quantitative
nation of exterior orientation include misidentified terrain data, covering a time span of more than 50
or mistyped control points. Fortunately, these years. Developments in digital photogrammetry
errors give rise to large residuals at the block allowed the acquisition of much denser DEMs
bundle adjustment stage and, if data redundancy (Brunsden & Chandler 1996). Following the
is high, are normally readily identifiable. In the efforts by Brunsden & Chandler in their studies
context of DEM generation a measure of internal on the Black Ven landslide (Chandler & Brunsden
reliability can be derived by comparing two 1995; Brunsden & Chandler 1996), the use of
DEMs of the same area but extracted from different multi-temporal DEMs has become more widely
stereopairs (Butler et al. 1998). DEMs can provide a adopted in landslide research (e.g. Cheng 2000;
perspective view of the area from any specified pos- Adams & Chandler 2002; Gentili et al. 2002; Van
ition. In combination with an orthophoto realistic Westen & Getahun 2003; Ager et al. 2004; Bitelli
views can be created, which are useful in analyses et al. 2004).
(Fig. 3). Figure 4 shows a ‘DEM-of-difference’ of the
central part of the Mam Tor landslide, created by
Multi-temporal DEMs subtracting DEMs of 1990 and 1973. The change
in elevation is draped over a standard DEM for a
DEMs not only provide a useful tool to enhance better interpretation. Dark areas represent a lower-
data analysis by perspective viewing, but also ing in elevation, whereas bright areas depict an
contain quantitative topographical data. Subtracting increase in height.

Fig. 4. On the right a ‘DEM-of-difference’ of the central part of the Mam Tor landslide, created by subtracting DEMs
of 1990 and 1973; the elevation change is draped over a standard DEM for better interpretation. Left image is an
orthophoto of the same area.
60 J. WALSTRA ET AL.

Orthophotos relevance of a map makes orthophotos particularly

valuable for Earth scientists (Chandler 2001).
Orthophotos combine the image characteristics of a Several researchers have shown the use of multi-
photograph with the geometric qualities of a map. temporal orthophotos to map horizontal surface
Unlike normal aerial photographs, relief displace- displacements. Powers et al. (1996) measured
ment is removed so that all ground features are movements of the Slumgullion landslide by deter-
displayed in their true ground position. This mining the displacement of surface features, such
allows the direct measurement of distances, areas, as trees and rocks, between two orthophotos
angles and positions. Orthophotos are created acquired at different times. Gentili et al. (2002)
through differential rectification, which eliminates measured the displacements of building corners
image displacements caused by photographic tilt on the Corniglio landslide, Italy, from orthophotos.
and terrain relief. The rectification procedure
requires a photograph with known orientation
parameters and a DEM. The collinearity concept Automatic extraction of displacement
can be used to determine the corresponding photo vectors
co-ordinates of all DEM points (Wolf & Dewitt
2000). The geometric quality of orthophotos is Digital techniques allow the potential of automatic
dependent on its source data; that is, the original measurement of objects on images. Kääb &
photographs, the functional model that relates Vollmer (2000) used an area-based cross-
photo to ground co-ordinates, and the quality of correlation algorithm to automatically map the
the DEM (Krupnik 2003). Hence, the quality con- velocity field of a rock glacier in the Swiss Alps
trols are similar to those for DEMs. The minimum from multi-temporal, orthorectified photographs.
grid spacing is bounded by the resolution of the The high density and accuracy of the velocity
original photographs, as a higher resolution would data provided by the technique make it possible
imply over-sampling. to extract meaningful strain-rate information. In
The combination of interpretative capabilities related studies their approach was successfully
of the original photographs with the positional applied to other types of superficial movements,

Fig. 5. A sequence of orthophotos obtained from different epochs, showing the progressively changing terrain surface
in the central part of the Mam Tor landslide.
 PHOTOGRAPHY FOR LANDSLIDE MONITORING 61

such as those of glaciers and rockslides (Kääb Conclusion

2002).
Nevertheless, some researchers have indicated This review has demonstrated the valuable infor-
that the accuracy of displacement vectors is mation, qualitative as well as quantitative, that is
limited by the relatively poor DEMs used in the captured by aerial photographs. It has also been
orthorectification process. Casson et al. (2003) shown that these data can be extracted relatively
developed an alternative approach, allowing the easily, using established techniques, and are well
creation of better DEMs than by most commercial suited for use in the monitoring stage of landslide
software packages. Their approach corrects for assessments. API reveals qualitative information
topographic distortion by estimating local slopes, on surface characteristics, which is very helpful
to improve the image matching performance. in detecting landslide features and in formulating
Kaufmann & Ladstädter (2002, 2004) used the statements about the mechanisms involved. Photo-
concept of pseudo-orthophotos, which, in combi- grammetrically derived products can provide
nation with a rough DEM, still contain the same quantification of the inferred processes: DEMs
stereo-information as the original photos, thus provide detailed data on surface topography,
allowing strict 3D reconstruction. Pseudo- whereas orthophotos can be used to measure
orthophotos are better suited for matching than surface displacements. Automated methods allow
original photo scans, as perspective distortions very dense and accurate data to be collected. In
have been removed to a great extent. An additional this way, the photographic archive can provide
advantage of this approach is that the obtained invaluable data on landslide evolution, thus
vectors are 3D, rather than horizontal. leading to a better understanding of landslide
Orthophotos were created from all photographic mechanisms.
epochs of Mam Tor (Fig. 5). Some clear features
can be identified throughout the entire series, and
can be manually measured to obtain displacement References
vectors. The mean displacement of the landslide
A DAMS , J. C. & C HANDLER , J. H. 2002. Evaluation of
over the period from 1953 to 1990 is 0.32 m a21, Lidar and medium scale photogrammetry for detecting
varying from 0.11 m a21 at the toe to 0.81 m a21 soft-cliff coastal change. Photogrammetric Record, 17
in the central, most active part. These values are 405– 418.
of comparable size to movement rates found by A GER , G., M ARSH , S., H OBBS , P., C HILES , R., H AYNES ,
Rutter et al. (2003), 0.04– 0.35 m a21 over the last M., T HURSTON , N. & P RIDE , R. 2004. Towards
century and up to 0.50 m a21 in recent years. Auto- automated monitoring of ground instability along
matic measurements can provide much denser dis- pipelines. In: Terrain and Geohazard Challenges
placement vector fields, allowing more detailed Facing Onshore Oil and Gas Pipelines. Thomas
comparison with the alternative surveys. However, Telford Ltd., London.
A L -D ABBAGH , T. H. & C RIPPS , J. C. 1987. Data sources
in the case of the Mam Tor images, differences for planning: geomorphological mapping of landslides
in image quality, illumination conditions and in north-east Derbyshire. In: C ULSHAW , M. G., B ELL ,
vegetation cover hampered this procedure. F. G., C RIPPS , I. C. & O’H ARA , M. (eds) Planning
and Engineering Geology. Engineering Geology
Special Publication 4, Geological Society, London,
101– 114.
Anonymous 1972. The preparation of maps and plans
Further research in terms of engineering geology. Report by the
Geological Society Engineering Group Working
The applied techniques have the ability to provide
Party. Quarterly Journal of Engineering Geology, 5,
both qualitative and quantitative information on 297– 367.
landslide movements from aerial photographs. B ALTSAVIAS , E. P. 1999. On the performance of
Continuing research aims at a thorough analysis photogrammetric scanners. In: F RITSCH , D. &
of their extra value to landslide investigations in S PILLER , R. (eds) Photogrammetric Week ’99,
terms of photographs available, techniques avail- Wichmann Verlag, Heidelberg, Stuttgart, 155– 173.
able and the usefulness of the products (type and B ITELLI , G., D UBBINI , M. & Z ANUTTA , A. 2004. Terres-
quality) in the analysis of landslide mechanisms. trial laser scanning and digital photogrammetry tech-
Further research will focus on how the different niques to monitor landslide bodies. International
datasets can be integrated to allow a comprehensive Archives of Photogrammetry and Remote Sensing,
XXXV(B5), 246–251.
analysis of the evolution of a landslide. Eventually, B RUNSDEN , D. & C HANDLER , J. H. 1996. Development
the techniques that have been developed at Mam of an episodic landform change model based upon
Tor will be validated in a case study in South the Black Ven mudslide, 1946–1995. In: A NDERSON ,
Wales, which should confirm their practical M. G. & B ROOKS , S. M. (eds) Advances in Hillslope
applicability. Processes. Wiley, Chichester, 869–896.
62 J. WALSTRA ET AL.

B RUNSDEN , D., D OORNKAMP , J. C., F OOKES , P. G., D E G RAAFF , L. W. S., D E J ONG , M. G. G., R UPKE , J. &
J ONES , D. K. C. & K ELLY , J. M. H. 1975. Large V ERHOFSTAD , J. 1987. A geomorphological mapping
scale geomorphological mapping and highway engin- system at scale 1:10 000 for mountainous areas. Zeits-
eering design. Quarterly Journal of Engineering chrift für Geomorphologie, 31, 229– 242.
Geology, 8, 227– 530. D RURY , S. A. 1987. Image Interpretation in Geology.
B UCKLEY , S. J. 2003. A geomatics data fusion technique Allen & Unwin, London.
for change monitoring PhD thesis, University of E CKHARDT , A., B RAUNECKER , B. & S ANDAU , R. 2000.
Newcastle upon Tyne. Performance of the imaging system in the LH Systems
B UTLER , J. B., L ANE , S. N. & C HANDLER , J. H. 1998. ADS40 airborne digital sensor. International Archives
Assessment of DEM quality for characterizing surface of Photogrammetry and Remote Sensing, XXXIII(B),
roughness using close range digital photogrammetry. 104–109.
Photogrammetric Record, 16, 271– 291. F RYER , J. G., C HANDLER , J. H. & C OOPER , M. A. R.
C ARRARA , A., C ARDINALI , M., G UZZETTI , F. & 1994. On the accuracy of heighting from aerial photo-
R EICHENBACH , P. 1995. GIS-based techniques for graphs and maps: implications to process modellers.
mapping landslide hazard. In: C ARRARA , A. & Earth Surface Processes and Landforms, 19,
G UZZETTI , F. (eds) Geographical Information 577–583.
Systems in Assessing Natural Hazards. Kluwer, G ENTILI , G., G IUSTI , E. & P IZZAFERRI , G. 2002. Photo-
Dordrecht, 135– 176. grammetric techniques for the investigation of the
C ASSON , B., D ELACOURT , C., B ARATOUX , D. & Corniglio landslide. In: A LLISON , R. J. (ed.) Applied
A LLEMAND , P. 2003. Seventeen years of the ‘La Geomorphology. Wiley, Chichester, 39–48.
Clapière’ landslide evolution analysed from ortho- G RANSHAW , S. I. 1980. Bundle adjustment methods
rectified aerial photographs. Engineering Geology, in engineering photogrammetry. Photogrammetric
68, 123 –139. Record, 10, 181–207.
C HANDLER , J. H. 1989. The acquisition of spatial data H AYDEN , R. S. 1986. Geomorphological mapping. In:
from archival photographs and their application to S HORT , S. & B LAIR , R. W. (eds) Geomorphology
geomorphology. PhD thesis, The City University, from Space. NASA. World Wide Web Address:
London. http://daac.gsfc.nasa.gov/geomorphology/GEO_11/
C HANDLER , J. H. 1999. Effective application of auto- GEO_CHAPTER_11.HTML (accessed 18 April
mated digital photogrammetry for geomorphological 2006).
research. Earth Surface Processes and Landforms, H OTTIER , P. 1976. Accuracy of close range analytical sol-
24, 51– 63. utions. Photogrammetric Engineering & Remote
C HANDLER , J. H. 2001. Terrain measurement using Sensing, 42, 345–375.
automated digital photogrammetry. In: G RIFFITHS , K ÄÄB , A. 2002. Monitoring high-mountain terrain defor-
J. S. (ed.) Land Surface Evaluation for Engineering mation from repeated air- and spaceborne optical data:
Practice. Engineering Geology Special Publication 18, examples using digital aerial imagery and aster data.
Geological Society, London, 13–18. ISPRS Journal of Photogrammetry & Remote
C HANDLER , J. H. & B RUNSDEN , D. 1995. Steady state Sensing, 57, 39– 52.
behaviour of the Black Ven mudslide: the application K ÄÄB , A. & V OLLMER , M. 2000. Surface geometry,
of archival analytical photogrammetry to studies of thickness changes and flow fields on creeping moun-
landform change. Earth Surface Processes and Land- tain permafrost: automatic extraction by digital
forms, 20, 255– 275. image analysis. Permafrost and Periglacial Processes,
C HANDLER , J. H. & C LARK , J. S. 1992. The archival 11, 315–326.
photogrammetric technique: further application K ALAUGHER , P. G., G RAINGER , P. & H ODGSON , R. L. P.
and development. Photogrammetric Record, 14, 1987. Cliff stability evaluation using geomorphologi-
241– 247. cal maps based upon oblique aerial photographs.
C HANDLER , J. H. & C OOPER , M. A. R. 1989. The extrac- In: C ULSHAW , M. G., B ELL , F. G., C RIPPS , I. C. &
tion of positional data from historical photographs and O’H ARA , M. (eds) Planning and Engineering
their application to geomorphology. Photogrammetric Geology. Engineering Geology Special Publication 4,
Record, 13, 69– 78. Geological Society, London, 163 –170.
C HENG , H.-H. 2000. Photogrammetric digital data K AUFMANN , V. & L ADSTÄDTER , R. 2002. Monitoring of
processing of Tsau-Lin big landslide. In: Proceedings active rock glaciers by means of digital photogram-
21st Asian Conference on Remote Sensing, Taipei, metry. International Archives of Photogrammetry
Taiwan, http://www.gisdevelopment.net/aars/acrs/ and Remote Sensing, XXXIV(3B), 108–111.
2000/ps1/ps102.asp. K AUFMANN , V. & L ADSTÄDTER , R. 2004. Documen-
C OLWELL , R. N. 1960. Manual of Photographic tation of the movement of the hinteres langtalkar
Interpretation. American Society of Photogrammetry, rock glacier. International Archives of Photogram-
Washington, DC. metry and Remote Sensing, XXXV(B7), 893–898.
C OOPER , M. A. R. & C ROSS , P. A. 1988. Statistical con- K RUPNIK , A. 2003. Accuracy prediction for ortho-image
cepts and their application in photogrammetry and generation. Photogrammetric Record, 18, 41– 58.
surveying. Photogrammetric Record, 12, 637–663. L ANE , S. N., J AMES , T. D. & C ROWELL , M. D. 2000.
C ROZIER , M. J. 1984. Field assessment of slope instabil- Application of digital photogrammetry to complex
ity. In: B RUNSDEN , D. & P RIOR , D. B. (eds) Slope topography for geomorphological research. Photo-
Instability. Wiley, Chichester, 103– 142. grammetric Record, 16, 793–821.
 PHOTOGRAPHY FOR LANDSLIDE MONITORING 63

L ILLESAND , T. M. & K IEFER , R. W. 1994. Remote S CHENK , A. F. 1996. Automatic generation of DEMs. In:
Sensing and Image Interpretation. Wiley, New York. C ARY , T., J ENSEN , J. & N YQUIST , M. (eds) Digital
L O , C. P. 1976. Geographical Applications of Aerial Photogrammetry. An Addendum to the Manual of
Photography. David & Charles, Newton Abbot. Photogrammetry. American Society for Photogram-
M ANTOVANI , F., S OETERS , R. & V AN W ESTEN , C. J. metry and Remote Sensing, Bethesda, MD, 247–250.
1996. Remote sensing techniques for landslide S KEMPTON , A. W., L EADBEATER , A. D. & C HANDLER ,
studies and hazard zonation in Europe. Geomorphol- R. J. 1989. The Mam Tor landslide, North Derbyshire.
ogy, 15, 213–225. Philosophical Transactions of the Royal Society of
N ORMAN , J. W., L EIBOWITZ , T. H. & F OOKES , P. G. London, Series A, 329, 503– 547.
1975. Factors affecting the detection of slope instabil- S OETERS , R. & V AN W ESTEN , C. J. 1996. Slope instabil-
ity with air photographs in an area near Sevenoaks, ity recognition, analysis, and zonation. In: T URNER ,
Kent. Quarterly Journal of Engineering Geology, 8, A. K. & S CHUSTER , R. L. (eds) Landslides Investi-
159–176. gation and Mitigation. National Academy Press,
P ARISE , M. 2003. Observation of surface features on an Washington, DC, 129– 177.
active landslide, and implications for understanding V AN W ESTEN , C. J. 1993. GISSIZ Training Package for
its history of movement. Natural Hazards and Earth Geographic Information Systems in Slope Instability
System Sciences, 3, 569–580. Zonation. Part 1: Theory. ITC, Enschede.
P OWERS , P. S., C HIARLE , M. & S AVAGE , W. Z. 1996. A V AN W ESTEN , C. J. & G ETAHUN , F. L. 2003. Analyzing
digital photogrammetric method for measuring the evolution of the Tessina landslide using aerial
horizontal surficial movements on the Slumgullion photographs and digital elevation models. Geo-
earthflow, Hinsdale Co., Colorado. Computers & morphology, 54, 77– 89.
Geosciences, 22, 651– 663. V AN W ESTEN , C. J., S EIJMONSBERGEN , A. C. &
R IB , H. T. & L IANG , T. 1978. Recognition and identifi- M ANTOVANI , F. 1999. Comparing landslide hazard
cation. In: S CHUSTER , R. L. & K RIZEK , R. J. (eds) maps. Natural Hazards, 20, 137–158.
Landslides Analysis and Control. National Academy V AN Z UIDAM , R.A. 1985. Aerial Photo-Interpretation in
of Sciences, Washington, DC, 34– 80. Terrain Analysis and Geomorphological Mapping.
R UTTER , E. H., A RKWRIGHT , J. C., H OLLOWAY , R. F. & Smits, The Hague.
W AGHORN , D. 2003. Strains and displacements W ALTHAM , A. C. & D IXON , N. 2000. Movement of the
in the Mam Tor landslip, Derbyshire, England. Mam Tor landslide, Derbyshire, UK. Quarterly
Journal of the Geological Society, London, 160, Journal of Engineering Geology and Hydrogeology,
735–744. 33, 105– 123.
S AVIGEAR , R. A. G. 1965. A technique of morphological W OLF , P. R. & D EWITT , B. A. 2000. Elements of Photo-
mapping. Annals of the Association of American grammetry, with Applications in GIS. McGraw–Hill,
Geographers, 55, 514–538. Boston, MA.

View publication stats