Full text of DOI:10.1111/1467-9671.00091

Transactions in GIS, 2002, 6(1): 5±15
Research Article
Automated Digital Photogrammetry:
A Valuable Tool for Small-scale
Geomorphological Research for the
Non-photogrammetrist?
Bruce Lascelles
Penny Anderson Associates Ltd
Buxton, Derbyshire
David Favis-Mortlock
School of Geography
Queen's University of Belfast
Tony Parsons
Department of Geography
University of Leicester
John Boardman
Environmental Change Institute
and School of Geography
University of Oxford
Abstract
Digital photogrammetry provides a tool with which to automatically generate
digital elevation models (DEMs). The necessary equipment is now both readily
available and affordable: thus there is considerable potential for this technique to be
widely adopted in geomorphological studies. But is it possible for geomorphologists
without a background in photogrammety to use it successfully? As part of a larger
study into rill initiation by overland flow, a non-metric digital camera and ERDAS
IMAGINE OrthoMAX software were used to generate small-scale DEMs of soil
surface microtopography. This paper reports on the procedure used, highlights
potential pitfalls, and comments on the quality of the resultant DEMs. Whilst
acquisition of high-quality images using a digital camera is relatively straight-
forward, problems were subsequently encountered due to the small size of the
internal imager and the need for camera calibration. Potential stumbling blocks in
the use of the software lay in the setting-up of ground control points and the use of
tie-points and check-points, as well as several software glitches not identified in the
current manual. Nonetheless, once these problems were overcome the technique
proved to be a simple, effective and fast tool for generating high quality microtopo-
graphical DEMs. This methodology shows great promise for future geomorpho-
logical studies that require these kinds of surface data.
Address for correspondence: Bruce Lascelles, Penny Anderson Associates Ltd, 60 Park Road,
Buxton, Derbyshire SK17 6SN, United Kingdom. E-mail: Bruce_Lascelles@paa-ecol.demon.co.uk
ß 2002 Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK and
3
50 Main Street, Malden, MA 02148, USA.

6
B Lascelles, D Favis-Mortlock, T Parsons and J Boardman
Introduction
1
Digital photogrammetry ostensibly provides an automated tool to generate high
density digital elevation models (DEMs). Recent developments in the field of
photogrammetry, particularly in terms of the software (often in a user-friendly
package) and hardware currently available, would appear to have opened up this
technique for use in geomorphology. The testing of geomorphological theory and the
modelling of processes increasingly requires high resolution quantitative data of
topography and topographic change at a wide range of scales (Stojic et al. 1998).
Already, the development and increasing availability of digital photogrammetry has
allowed new applications of photogrammetry for topographical determination (Fryer
et al. 1994, Brunsden and Chandler 1996, Dixon et al. 1998, Pyle et al. 1997, Stojic et al.
1
998, Stojic and Shannon 1998). A major advantage of automated digital
photogrammetry over other surveying techniques is the greatly increased rates at
which DEMs of consistent precision can be derived. This may be up to 1000 times
faster when compared to the use of a modern Total Station or digital tacheometer
(
Chandler 1999).
As a complement to this position, recent developments in computational models to
monitor and predict hydrology, erosion, and landscape evolution at a similarly wide
range of scales place heavy reliance on the availability and integrity of DEMs (Walker
and Willgoose 1999), making their acquisition for geomorphological studies highly
desirable. These advances have led to an upsurge in interest in photogrammetric
methods within the geomorphological research community (Chandler 1999). However,
results from these applications of photogrammetry are only slowly starting to appear in
the literature (e.g. Fryer et al. 1994; Brunsden and Chandler 1996, Dixon et al. 1998,
Pyle et al. 1997, Butler et al. 1998, Stojic et al. 1998, Stojic and Shannon 1998, Lane
2
000).
Nonetheless, the techniques employed in the use of photogrammetry to produce
DEMs are not trivial: thus there is also considerable potential for problems to occur. As
Chandler (1999) has suggested, it would be undesirable if problems that are really due
to inexperience and over-expectation were confused with the science of
photogrammetry itself. For example, Walker and Willgoose (1999) suggest that
geomorphologists, among others, are not fully aware of the limitations of DEMs as
sources of spatial information. Although there are some published reports on the
accuracy of DEM-derived elevations and slopes in the literature (Cooper and Cross
1
988, Lee et al. 1992, Bolstad and Stowe 1994, Robinson 1994, Heritage et al. 1998,
Wise 1998, Gooch et al. 1999, Walker and Willgoose 1999), DEMs are created,
distributed, and used very frequently without any reference to the possible error or to
Â
methods applied to its disclosure or correction (Felicõsimo 1994, Wise 1998). Walker
and Willgoose (1999) present a method for improving photogrammetric DEMs in
relation to ground-truthed data, but suggest that research into a large number of DEMs
of different terrain types and different grid sizes is required before any
recommendations can be made regarding post-processing requirements.
Of the small quantity of geomorphological research using digital photogrammetry
actually published, there appear to be few instances where the technique has been
applied for objects as small as several metres (Pyle et al. 1997, Butler et al. 1998, Stojic
et al. 1998, Chandler 1999). Chandler (1999) points out that, although digital
photogrammetric software is widely available to the research community, it has
ß Blackwell Publishers Ltd. 2002

Automated Digital Photogrammetry
7
primarily been developed for remotely sensed imagery and vertical aerial photography.
However, Stojic et al. (1998) and Chandler (1999) demonstrate that such software
could function at much larger photo-scales. The current availability of affordable high-
resolution digital cameras means that digital photogrammetry might also be applied to
close-range measurement problems, particularly change detection and the
quantification of the impact of micro-scale geomorphological processes covering areas
2
of 1±400 m . Of the three papers that report on the use of digital photogrammetry at
these scales, two present some quantitative assessment of DEM quality in terms of
either precision or accuracy. Pyle et al. (1997) report an accuracy of approximately 1/
1
000 of the camera/object distance, a figure that is towards the lower end of accuracies
achieved in aerial photography.
With the rapid advances still taking place (e.g. the development of OrthoBASE
by ERDAS) it is likely that more and more use will be made of the technique by
geomorphologists. But is it a reliable and straightforward technique for geomorphol-
ogists with no background in photogrammetry? This paper aims to build on the
advice presented by Chandler (1999) on the effective application of automated
digital photogrammetry for geomorphological research. An example is presented of
the use of a non-metric digital camera and the ERDAS IMAGINE OrthoMAX
2
software by a non-photogrammetrist for deriving DEMs of a 7 m soil surface.
Potential pitfalls are highlighted and the value of DEMs generated by this type of
study is assessed.
2
Methodology
Digital photogrammetry was used within a larger project that aimed to study rill
initiation by overland flow and its role in modelling soil erosion. DEMs of the soil
surface before and following the application of simulated rainfall were required in
order to validate a dynamic, process-based model of rill initiation and development
(RillGrow) based upon the notion of emergent self-organisation (Favis-Mortlock 1996,
1
998; Favis-Mortlock et al. 1998, 2000). Simulated rainfall was applied to various soil
surfaces set at different slope angles. Surface flow, discharge, sediment removal and
flow velocities were all measured during 30 minute experimental runs. The soil was
held within a dual axis soil slope (flume), 4 Â 1.75 Â 0.15 m. For each experiment, a soil
surface with pre-specified characteristics was created; these included smooth (raked
and rolled), raked, and undulating. Prior to and following each experimental run,
overlapping images of the soil surface were acquired to produce stereo-pairs necessary
to create DEMs. The methodology involved in this process, and the problems
encountered, are outlined in the following sections.
2
.1 Ground Control and Tie Points
The flume sides form a 4 Â 1.75 m fixed rectangle. Along both long edges of the flume
0 ground control points (GCPs) were marked with dots, 2±3 mm in diameter, of white
2
waterproof paint. The location of these relative to the upper left-hand corner of the soil
surface within the flume (denoted as x ˆ 0, y ˆ 0) was measured using a steel ruler.
Chandler (1999) suggests that a total station, using the intersection method, should be
employed for close range or laboratory-based location of GCPs. However, in the
ß Blackwell Publishers Ltd. 2002

8
B Lascelles, D Favis-Mortlock, T Parsons and J Boardman
present case it was possible to record the distance between the centres of each control
point on the fixed-dimension flume using a steel ruler with 0.5mm graduations; the
angle between flume walls being exactly 90ë. The use of a total station was therefore
not deemed necessary. These GCPs could therefore be given x and y co-ordinates. The
flume was regularly checked, on all sides, with a spirit level to ensure that when the
slope was lowered in order to image the surface, the flume was level (the flume being
levelled using four adjustable jacks). Thus, the GCPs lying along the two long edges of
the flume were assigned a z (height) value of 0. These points, therefore, define the
datum.
Other points on the flumes' structure that could easily be identified in overlapping
images were also located and their co-ordinates noted, again relative to the upper left-
hand corner of the soil surface and the designated datum. These all had positive heights
relative to the datum; points on the floor that could be identified in overlapping images
were used to provide GCPs with negative heights in relation to the datum. Because of
the nature of the experimental set-up it was not possible to fix permanent control
points across the soil surface. It is suggested by ERDAS (Technical Support, pers.
comm., 1999) that GCPs distributed only around the edge of the area being studied can
result in poor modelling of lens distortion and so poor DEM acquisition over the
central portion of the surface. To compensate for this, when locating the GCPs in the
OrthoMAX software, tie points were also measured. In this instance, these were
features on the soil surface (usually small pieces of quartz) that could be located in
overlapping images. Such features were measured in all the images in which they
occurred, to enhance the triangulation process across the surface that would otherwise
be devoid of ground control. Note that the operator does not need to know any of the
co-ordinates of the feature used for each tie point, this is determined during the
triangulation procedure.
2
.2 Camera Calibration
A Kodak DCS420 digital camera, with a 24 mm Nikkor lens attached, was used to
acquire the images. This is a non-metric `off-the-shelf' camera, requiring calibration to
acquire the interior parameters (focal length offset (Á f), principal point offset (x , y ),
p
p
radial distortion parameters (k , k , k ), and decentring distortion parameters (p , p ))
1
2
3
1
2
applicable to this camera. However, there are several important issues to take into
consideration when calibrating the DCS420 digital camera.
It is clearly undesirable for the CCD (charge coupled device) that captures the
image in the digital camera to move with respect to the camera lens, as it is thought to.
It has been suggested that any movement will be negligible, but some workers (Mark
Nearing, pers. comm., 1999) are taking the precaution of having the CCD chip firmly
fixed in the cameras they are using. Obviously, movement of the CCD chip will alter
any calibration parameters already defined for the camera. Chandler (pers. comm.,
1
999) suggested that it would be possible to avoid problems related to movement in the
CCD device by carrying out a self-calibrating bundle adjustment on every overlapping
pair of images, so generating a calibration for each stereo-pair.
However, this procedure is time-consuming and must be repeated many times.
With little firm information available on this issue it was therefore decided to calibrate
the camera once only and to use these settings throughout the project. Thus, this
approach assumes no movement of the CCD during the experiments; although no tests
ß Blackwell Publishers Ltd. 2002

Automated Digital Photogrammetry
9
to check this were undertaken here and it is an area that needs further investigation.
Camera calibration was carried out by the Optical Meteorology Centre at City
University. It is important to note here that if a single calibration is undertaken, the
calibration parameters apply only to the camera in the state it was in when the
calibration was performed. Change in focus or removal and re-attachment of the lens
will alter the parameters.
Another related issue to be aware of, certainly for the non-photogrammetrist, is
that the camera radial distortion parameters (k , k , k ), supplied as part of the camera
1
2
3
calibration, are not the same as the polynomials (k , k , k ) required by the camera
0
1
2
model in OrthoMAX. Using the radial distortion parameters radial distortions at
various radial distances must to be computed: it is these that can then be input into the
camera model in OrthoMAX. The polynomials required are then computed within
OrthoMAX.
2
.3 Camera Set-up
A limitation of using the DSC420 camera is that the CCD imager has dimensions of
1
3.8 Â 9.2 mm, comprising 1524 Â 1012 pixels. Thus the CCD provides one quarter of
the image area that a 35 mm film provides, and so a standard AF 24 mm Nikkor lens
provides the same coverage as a 50 mm lens. This required the camera to be mounted
4
m above the soil surface so that the whole width of the surface and the outlying GCPs
could be covered by one image. A lower camera height would have increased the
number of images required, resulting in longer processing times and larger computer
storage requirements. Use of oblique photography would have reduced the `flying
height' required, but due to the additional (and potentially confusing) problems this
adds during data processing (Chandler 1999) this approach was not adopted. The
camera was attached to a metal pole using the normal camera screw mount, and this
was adjustably attached to a 4.5 m long aluminium pole. This allowed the camera to be
set up and then moved into position from the floor.
2
.4 Creating DEMs
The standard sequence of events, as outlined in the OrthoMAX manual, was
undertaken to turn the overlapping images into DEMs. The default settings given by
OrthoMAX in the DEM tool were used.
There are, however, a number of points that could potentially cause problems.
Some of these result from glitches or bugs in the software that have not been mentioned
in the manual. Of particular importance here were `steps' in elevation that occurred
between DEMs when they were mosaiced together (using the mosaic tool in
IMAGINE). The removal of these aberrations prior to the use of the DEM in
modelling is vital; if they remain model outcomes will be influenced and so the value of
any model validation process will be reduced. There are many reasons why these
artefacts may occur; through lens distortion, inaccuracies in camera exterior
orientation or inaccuracies in the input and location of control points. However, if
steps are still created when no further improvement in the quality of input data is
possible, one fairly effective approach is to mosaic the two adjoining DEMs first, and
then to mosaic the resulting image to the remainder of the DEMs. If the sequence
outlined here is followed, then when the operator is familiar with assessing the
ß Blackwell Publishers Ltd. 2002

10
B Lascelles, D Favis-Mortlock, T Parsons and J Boardman
triangulation results and locating possible sources of error, the process of producing a
single DEM from multiple overlapping images is reasonably straightforward.
3
Assessing DEM Quality
Chandler (1999) stresses the importance of establishing the required quality of any
measured data, particularly the accuracy of the DEM, before carrying out any
photogrammetric measurement. The accuracy required of course depends on the use
to which the DEM is to be put. Richards et al. (1995) suggest that the accuracy of a
DEM used to generate a simple contour plot for qualitative purposes can be far
lower than a DEM used for process-based simulation modelling, as in this project. It
was expected, from image scale, that an accuracy of Æ0.5 mm should be achievable.
Chandler (1999) describes briefly some of the methods that can be used to assess the
quality of generated DEMs produced for geomorphological research. He suggests
that the convention used in surveying and photogrammetry be adopted, in which
overall data quality is described by quantifying three components: precision,
accuracy, and reliability.
The precision of the derived data is a function of the precision of the photo
measurements and the scale and geometry of the photographs. Numerical estimates of
precision can be obtained through examination of the variance propagation output
from OrthoMAX (Chandler 1999). Accuracy of data is dependent upon the accuracy of
the photo-control and knowledge of the internal geometry of the camera. Here,
Chandler (1999) suggests that independent check-points are essential to quantify the
accuracy of a DEM, by comparing the positions of the check points with their
estimated co-ordinates and computing the root mean square value. However, Pyle et al.
(
1997) suggest that the provision of ground-truthed data is likely to offer only a partial
assessment of DEM quality in small-scale work. They highlight factors such as the
difficulty of measuring more than a small number of points, ensuring survey accuracies
better than a few millimeters, and unambiguously identifying each location with the
corresponding point on the DEM.
At the start of this project it was assumed that check points would be used to assess
accuracy, and that these would be in the form of GCPs that were not used in the
triangulation process (i.e. around the edges of the flume). However, despite apparent
good (qualitatively) DEM restitution over the soil area, OrthoMAX failed in all cases
to produce realistic elevations around the edges of the flume, where the GCPs were
located. It was not therefore possible in practice to make use of these independent
check points to assess accuracy; in future experimental work it would be prudent to
include GCPs within the central portion of the flume to ensure accuracy tests could be
undertaken. However, it was possible to carry out the test for reliability, and therefore
potentially accuracy, suggested by Stojic et al. (1998). Overlapping areas of two
adjacent DEMs were subtracted from one another. Differences in elevation for an area
that should have the same elevation statistics can be represented by the mean and
standard deviation values for the overlapping area.
Chandler (1999) suggests that it might also be useful to use the notion of `external
reliability', as used in surveying, which measures the impact of an error upon derived
data, typically co-ordinates. DEMs produced using the method outlined have been used
as input into the RillGrow model. An assessment will also be made of the output from
ß Blackwell Publishers Ltd. 2002

Automated Digital Photogrammetry
11
RillGrow after perturbing a DEM slightly to test the influence of small errors on the
functioning of the model (cf. Burrough 1998).
4
Quality of DEMs and Their Value in a Process-based Model
4
.1 Qualitative Analysis of DEMs
Figure 1 shows two DEMs produced using the methodology outlined in this paper. The
top image is the soil surface at the start of the experiment, the bottom image is of the
surface after the application of simulated rainfall for 30 minutes with the flume slope
set to 15ë.
These images depict a smooth but undulating surface that was created for one
experimental run. Evidence of erosion can be seen in the form of discontinuous rills
that become more continuous towards the foot of the slope. They are discontinuous in
this instance due to erosion occurring mostly where flow has been forced to cut
through a raised area. Qualitative inspection of these DEMs with reference to the
original images of the surface showed that the DEMs produced were an excellent
Figure 1 (a) Initial surface topography; (b) Surface topography following application of
rs ai mn gu el a ti se d8 cr am infall for 30 minutes with a slope of 15. Area shown in 4 Â 1.76 m. Elevation
ß Blackwell Publishers Ltd. 2002

12
B Lascelles, D Favis-Mortlock, T Parsons and J Boardman
representation of the soil microtopography. All rills visible in the original images are
found in the DEMs
4
.2 Quantitative Assessment of DEM Quality
The precision of the DEMs was estimated from data generated in OrthoMAX. The
precision of the x, y and z co-ordinates for the left-hand DEMs depicted in Figure 2 are
0
.62 mm, 0.67 mm, and 0.1 mm. Precision estimates for the other DEMs in this and
other series were similar, indicating a consistency in the production of DEMs from
different blocks of imagery. Tie point precision (as shown by the `View Propagation'
capability in OrthoMAX) was similar; DEM precision is therefore good.
Using the method outlined by Stojic et al. (1998) to assess DEM accuracy in the
absence of check points, the overlapping areas of adjacent DEMs were subtracted. The
resultant DEM for the overlap area, as shown in Figure 2, had a mean elevation of
0
.05 mm and a standard deviation of 0.14 mm. These figures fall well within the
theoretically expected accuracy and compare very favourably with those published in
the literature for similar studies. Stojic et al. (1998) report the subtracted area in their
study having an elevation mean of 0.7 mm and a standard deviation of 0.6 mm,
Figure 2 Assessment of DEM accuracy using the overlapping areas of adjacent DEMs.
Overlap area is 1.76 Â 0.40 m
ß Blackwell Publishers Ltd. 2002

Automated Digital Photogrammetry
13
although the overall area they were interested in was nearly five times that used in this
study. Stojic et al. (1998) also report the existence of steps in elevation in the left and
right limits of each DEM that they attribute to a systematic error. In this study, steps of
no more than several millimetres remained in two DEMs and were not possible to
remove, but these occurred after mosaicing of several DEMs. No systematic errors
were apparent in the individual DEMs.
4
.3 Viewing and Manipulating DEMs
In the authors' opinion IMAGINE did not provide a user-friendly system for viewing
and manipulating the DEMs produced. DEMs were exported to IDRISI where they
could more easily be manipulated and, importantly for presentations, displayed in
colour. Although within IMAGINE it is possible to display DEMs in 3-D using
VirtualGIS, and in IDRISI using the ORTHO module, neither of these were adequate
for our needs. Presentation and manipulation of the DEMs produced, in photo-realistic
colour 3-D, was therefore carried out using the OpenGL-based software called GL-
Display produced by one of the authors (Favis-Mortlock). This reads the same DEM
file format as IDRISI.
4
.4 DEM Value in a Process-based Model of Small-scale Geomorphological
Processes
The photogrammetrically-derived DEMs were each used as an input for a simulation of
the RillGrow model (Favis-Mortlock 1996, 1998; Favis-Mortlock et al. 1998, 2000).
RillGrow's approach is to effectively discretise erosive flow by considering it to
comprise individual `packets' of runoff located at the nodes of the DEM grid.
Relatively simple quantitative rules govern the movement of water between packets,
and the erosive and depositional modification of the elevations of the nodes of the
microtopographic grid.
As a result of very many local (i.e. grid-scale) interactions between packets, the
system produces larger-scale `emergent' responses. These include the development of
realistic dendritic channel networks; hydrograph and sedigraph time series; and whole-
area responses. Results from an ongoing validation of the model are very encouraging,
and indicate that, given the original microtopography, RillGrow can plausibly predict
where erosive flow paths will develop on a given soil surface.
These larger-scale `emergent' responses (which we can validate) are thus the result
of myriad local-scale flow routings within the model (which we cannot validate). Since
local flow routing is driven entirely by differences in microtopographic elevation
between DEM grid nodes, successful validation of the model is sensitively dependent
on DEM quality. The model's good results in our validation thus imply high-quality
DEMs (Favis-Mortlock et al. 2000).
5
Conclusions
The methodology and analysis presented here may, in photogrammetric terms, be
simplistic. However, for a geomorphological researcher with no previous experience of
a technique for which there is a distinctly limited literature (and with the additional
ß Blackwell Publishers Ltd. 2002

14
B Lascelles, D Favis-Mortlock, T Parsons and J Boardman
stumbling blocks of software and documentation deficiencies) this technique could
present a significant challenge. Papers published within the last two years by Stojic et
al. (1998) and Chandler (1999) ± and now this study ± have sought to address this lack
of information with particular emphasis on inexperienced users. Certainly, the results
presented here are most encouraging. DEMs have been derived that are, both
qualitatively and quantitatively, good representations of the real surfaces. The value of
close-range non-metric digital photogrammetry and DEM acquisition using the
ERDAS IMAGINE and OrthoMAX software in geomorphological modelling studies is
confirmed by these results. However, there are still questions that remain unanswered.
Are the default settings in OrthoMAX for DEM acquisition optimum for close-range
work? Exactly what methods are best for assessing DEM quality in terms of the
requirements of geomorphological studies? And just what is achievable using this
technique?
Chandler (1999) is concerned that the photogrammetric process may become over-
trivialised as it becomes more widely used for such studies. There is certainly, from the
experience gained during this project, the possibility that inexperienced users may
become disillusioned with the science of photogrammetry when an overly ambitious
project fails to achieve the expected results in the given time-frame. However, this will
become less likely as more studies are published, and the methodology and its strengths
and limitations become better known, and as a standardised approach is evolved.
Methods of assessing accuracy and precision also require some standardisation, with
particular thought to their use by workers with little background in this science.
Acknowledgements
This work is part of NERC project GR3/11417: `Rill initiation by overland flow and its
role in modelling soil erosion' and was carried out in the Department of Geography at
the University of Leicester, and the Environmental Change Institute at the University of
Oxford. The authors are grateful for advice from, and discussions with, Jim Chandler
(
University of Loughborough, UK), Mark Nearing (USDA-ARS National Soil Erosion
Laboratory, West Lafayette, USA), and the ERDAS Technical Support Team whilst
undertaking this work. We also thank Xinchi Wang and Tim Clarke (City University,
UK) for undertaking and explaining the process of camera calibration.
References
Bolstad P V and Stowe T 1994 An evaluation of DEM accuracy: Elevation, slope, and aspect.
Photogrammetric Engineering and Remote Sensing 60: 1327±32
Brundsen D and Chandler J H 1996 The continuing evolution of the Black Ven mudslide, 1946±
9
5. In Brooks S and Anderson M (eds) Advances in Hillslope Processes. Chichester, John
Wiley and Sons: 869±98
Burrough P A 1998 Dynamic modelling and geocomputation. In Longley P A, Brooks S M,
McDonnell R, and Macmillan R (eds) Geocomputation: A Primer. Chichester, John Wiley
and Sons: 165±91
Butler J B, Lane S N, and Chandler J H 1998 Assessment for DEM quality for characterizing
surface roughness using close range digital photogrammetry. Photogrammetric Record
1
6(92): 271±91
Chandler
J
1999 Effective application of automated digital photogrammetry for
ß Blackwell Publishers Ltd. 2002

Automated Digital Photogrammetry
15
geomorphological research. Earth Surface Processes and Landforms 24: 51±63
Cooper M A R and Cross P A 1988 Statistical concepts and their application in photogrammetry
and surveying. Photogrammetric Record 12(71): 637±63
Dixon L F J, Barker R, Bray M, Farres P, Hooke J, Inkpen R, Merel A, Payne D, and Shelford A
1
998 Analytical photogrammetry for geomorphological research. In Lane S N, Richards K S,
and Chandler J H (eds) Landform Measurement, Modelling and Analysis. Chichester, John
Wiley and Sons: 63±94
Favis-Mortlock D T 1996 An evolutionary approach to the simulation of rill initiation and
development. In Abrahart R J (ed) Proceedings of the First International Conference on
GeoComputation (Volume 1). Leeds, School of Geography, University of Leeds: 248±81
Favis-Mortlock D T 1998 A self-organising dynamic systems approach to the simulation of rill
initiation and development on hillslopes. Computers and Geosciences 24: 353±72
Favis-Mortlock D T, Guerra A J T, and Boardman J 1998 A self-organizing dynamic systems
approach to hillside rill initiation and growth: Model development and validation. In
Summer W, Klaghofer E, and Zhang W (eds) Modeling Soil Erosion, Sediment Transport,
and Closely Related Hydrological Processes. Wallingford, International Association of
Hydrological Sciences Publication 249: 53±61
Favis-Mortlock D T, Boardman J, Parsons A J, and Lascelles B 2000 Emergence and erosion: A
model for rill initiation and development. Hydrological Processes 14: 173±205
Felicisimo A M 1994 Parametric statistical method for error detection in digital elevation models.
ISPRS Journal of Photogrammetry and Remote Sensing 49: 29±33
Fryer J G, Chandler J H, and Cooper M A R 1994 On the accuracy of heighting from maps and
aerial photographs: Implications for process modellers. Earth Surface Processes and
Landforms 19: 577±83
Gooch M J, Chandler J H, and Stojic M 1999 Accuracy assessment of digital elevation models
generated using the ERDAS IMAGINE OrthoMAX digital photogrammetric system.
Photogrametric Record 16(93): 519±31
Heritage G L, Fuller I C, Charlton M E, Brewer P A, and Passmore D P 1998 CDW
photogrammetry of low relief fluvial features: Accuracy and implications for reach-scale
sediment budgeting. Earth Surface Processes and Landforms 23: 1219±33
Lane S N, James T D, and Crowell M D 2000 Application of digital photogrammetry to complex
topography for geomorphological research. Photogrammetric Record 16(95): 793±821
Lee J, Snyder P K, and Fisher P F 1992 Modeling the effect of data errors on feature-extraction
from digital elevation models. Photogrammetric Engineering and Remote Sensing 58: 1461±
7
Pyle C J, Richards K S, and Chandler J H 1997 Digital photogrammetric monitoring of river
bank erosion. Photogrammetric Record 15(89): 753±64
Richards K S, Arnold N, Lane S, Chandra S, El Hames A, Mattikalli N, and Chandler J H 1995
Numerical landscapes: Static, kinematic and dynamic process-form relations. Zeitschrift fur
Geomorphologie, Supplementband 101: 201±20
Robinson G L 1994 The accuracy of digital elevation models derived from digitised contour data.
Photogrammetric Record 14(83): 805±14
Stojic M, Chandler J, Ashmore P, and Luce J 1998 The assessment of sediment transport rates by
automated digital photogrammetry. Photogrammetric Engineering and Remote Sensing 64:
3
87±95
Stojic M and Shannon J 1998 Digital cameras and their application to digital photogrammetry
for landform modelling. In Proceedings of the International Society for Photogrammetry
and Remote Sensing Symposium on Data Integrations, Systems, and Techniques,
Cambridge, 13±17 July, 1998: 306±11
Walker J P and Willgoose G R 1999 On the effect of digital elevation model accuracy on
hydrology and geomorphology. Water Resources Research 35: 2259±68
Wise S M 1998 The effect of interpolation errors on the use of digital elevation models in
geomorphology. In Lane S N, Richards K S, and Chandler J H (eds) Landform Monitoring,
Modelling, and Analysis. Chichester, John Wiley and Sons: 1±17
ß Blackwell Publishers Ltd. 2002