Geomorphology 137 (2012) 94106

Contents lists available at ScienceDirect

Geomorphology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g e o m o r p h

Geomorphometry and landform mapping: What is a landform?

Ian S. Evans
Department of Geography, Durham University, South Road, Durham City, DH1 3LE, UK

a r t i c l e

i n f o

Article history:
Accepted 8 September 2010
Available online 27 March 2011
Keywords:
Scaling
Allometry
Operational denitions
Landform delimitation
DEMs

a b s t r a c t
Starting from a concept of the land surface, its denition and subdivision from Digital Elevation Models
(DEMs) is considered. High-resolution DEMs from active remote sensing form a new basis for
geomorphological work, which is moving on from consideration of whether data are accurate enough to
how the surface of interest can be dened from an overabundance of data. Discussion of the operational
denition and delimitation of specic landforms of varying degrees of difculty, from craters to mountains, is
followed by the applicability of fuzzy boundaries. Scaling, usually allometric, is shown to be compatible with
the scale-specicity of many landforms: this is exemplied by glacial cirques and drumlins. Classication of a
whole land surface is more difcult than extraction of specic landforms from it. Well-dissected uvial
landscapes pose great challenges for areal analyses. These are tackled by the delimitation of homogeneous
elementary forms and/or land elements in which slope position is considered. The boundaries are mainly
breaks in gradient or aspect, but may also be in some type of curvature: breaks in altitude are rare. Elementary
forms or land elements are grouped together into functional regions (landforms) such as hill sheds. It may
often be useful to recognise fuzziness of membership, or core and periphery of a surface object.
Plains and etched or scoured surfaces defy most of these approaches, and general geomorphometry remains
the most widely applicable technique. It has been applied mainly within arbitrary areas, and to some extent to
drainage basins, but more experimentation with mountain ranges and other landforms or landform regions is
needed. Geomorphological mapping is becoming more specialised, and its legends are being simplied. Its
incorporation into geographical information systems (GIS) has required greater precision with denitions,
and the separation of thematic layers, so that it is converging with specic geomorphometry and becoming
more exible and more applicable, with a broader range of visualisation techniques.
2011 Published by Elsevier B.V.

1. Introduction

A denition is useless unless the thing dened can be recognized

in terms of the denition when it occurs. H. Jeffreys, 1961 p.8
Recent work has asked questions such as What (or where) is a
mountain? (Smith and Mark, 2003; Fisher et al., 2004), which raises
perceptual and linguistic as well as geomorphological considerations.
Here I tackle the question What is a landform?, and take a more
strongly geomorphology-centred approach. Landforms are central to
many denitions of geomorphology, e.g.:
Geomorphology, or the study of landforms (Lobeck, 1939 p.3)
Geomorphology is the study of landforms. (Pitty, 1971 p.1)
After about 1860 the study of landformswas later also known
as physiography or geomorphology. (Chorley et al., 1964 p. xi)
Geomorphology is the science that investigates the landforms of
the earth. (Ahnert, 1998 p.1).

E-mail address: I.s.evans@durham.ac.uk.

0169-555X/$ see front matter 2011 Published by Elsevier B.V.
doi:10.1016/j.geomorph.2010.09.029

Others put it slightly differently:

Geomorphology is the study of the Earth surface. (Butzer, 1976 p.1)

Geomorphology is the scientic study of the geometric features of
the Earth's surface. (Chorley et al., 1984 p.3)
Geomorphology is the science concerned with the form of the
landsurface and the processes which create it. (Summereld, 1991
p.3).

Recent research has concentrated on processes, materials and

chronology, but the land surface itself remains central in geomorphology.
Its analysis is essential in the verication or calibration of models (Dietrich
et al., 2003). Thus, the specialised subject of geomorphometry (Pike, 2000;
Hengl and Reuter, 2009) has grown tremendously in recent years as more
and better DEMs (Digital Elevation Models) have become available.
An important distinction between landform and land surface form
parallels the distinction between specic and general geomorphometry
(Evans, 1972; Goudie, 1990). Land surface form is continuous and covers
the whole globe (both subaerial and subaqueous), as well as other
planets, moons and asteroids. General geomorphometry analyzes this

I.S. Evans / Geomorphology 137 (2012) 94106

continuous eld. Landforms are bounded segments of a land surface and

may be discontinuous: they need not cover the whole surface (Fig. 1).
Specic geomorphometry analyzes the geometric and topological
characteristics of landforms. The question of what to do with the bits
left between deserves consideration.
The denition of landforms of various sorts is an essential part
of geomorphological mapping. Complete delimitation is a further
necessity for specic geomorphometry: to measure their geomorphological characteristics, individual landforms must be separated
from their surroundings. Characteristics such as area and mean slope
gradient require a complete boundary: thus, it is not sufcient to map
a cirque by a crescent (indicating the headwall crest in a generalised
way) as on many traditional geomorphological maps. A prerequisite
for specic geomorphometry is to draw a closed boundary: this is the
procedure known as delimitation, marking the limits of the landform
(Evans, 1987; 2010). The question whether these limits can be fuzzy,
or must be precise, is considered below.
This paper moves from the denition and properties of specic
landforms in relatively clear-cut cases, through more difcult situations,

95

to consider nally the application of general geomorphometry and

geomorphological mapping. First, the concept of the land surface is
dened, and the DEMs from which it may be analysed for landform
denition are considered. Cirques and drumlins are used as examples of
clearly-bounded landforms: other landforms and terrains vary in the
difculty of delimiting forms. The objectives are to make connections, to
review concepts and progress in dening and measuring landforms, and
to illustrate landform allometry and scale-specicity. Such a concern
with the essence of landforms relates to the philosophical eld of
ontology (Smith and Mark, 2003). Hierarchies and broader-scale
divisions, such as land systems, landscape types and physiographic
provinces, are not discussed here. The focus is on the geometry of
landforms, not topology or relation to materials and processes, and on
areal more than linear features.
2. The land surface concept: scale
Reecting perhaps the inuence of many years in a Geography
Department, my concept of the land surface starts at the human scale:

Fig. 1. Comparison of drumlins mapped for an area close to Loch Lomond in Scotland by (a) intensive eld mapping by J. Rose. Drumlins, crestlines and summits are mapped. The
esker (marked E) was erroneously mapped as two drumlins in (b), mapping from the NEXTMap 5 m DEM. In (b) solid black drumlins coincide with eld mapping; black outlines are
those found from the DEM but not in the eld, and the boxes highlight areas where eld mapping found examples that were not seen in the DEM.
From Clark et al. (2009).

96

I.S. Evans / Geomorphology 137 (2012) 94106

1.5 m. It does not end there, but a long tradition in slope studies
involves measuring over this slope length, notably with Pitty's (1971
pp. 402404; Cox, 1990) slope pantometer. Finer-scale variation is
described as microrelief, although its upper limit has rarely been
clearly dened. Whatever the pros and cons of generalising to 1 or 2
m, it does give a common basis from which measurements can be
compared.
Many geomorphologists have focused on the surface smoothed at
a scale of a few metres. In slope prole survey, Young (1972 p. 146)
recommended that No measured length shall be more than 20 m or
less than 2 m for topography of normal scale. Gerrard and
Robinson (1971) analysed gradients for xed measured lengths
from 1.5 m to 10 m, and discussed the effects of small protrusions and
depressions (microrelief) that can give variations of a few degrees for
measuring lengths of a few metres. Pitty (1971) advocated xed
lengths and used a frame giving a constant slope length of 1.5 m
(previously 5 feet, i.e. 1.52 m) for gradient measurement: this human
scale relates to the practicality of one-person operation. Debate
concerning xed versus variable measuring lengths along proles
continued (Cox, 1990) and had some parallels with the debate about
xed grids, adaptive grids and irregular triangulations as bases for
DEMs. In slope proling the basic idea was to exclude individual
particles (stones), minor bumps and depressions, and repeated
microrelief such as earth hummocks (best analysed separately, by
appropriate means of analysis, as are vertical cliffs, overhangs, pipes
and caves). The inuence of grain size was removed; boulders would
generally be avoided. Buildings and vegetation were also excluded.
Boulders several metres long, for example in rockfall or glacial
deposits, pose a problem. If they sit on the surface, they can be
excluded from the concept of land surface in the same way as trees or
buildings. If the boulders are close together or partly buried (and all
degrees of burial exist), however, it is more appealing to pass a surface
through the boulders, either averaged if the boulders are close
together, or interpolated from the surrounding surface if this is
relatively smooth. Thus, we reach a geomorphological concept of the
(real) land surface, with a lower limiting scale of a metre or two, and
relating to the ground, a continuous body or aggregate of material,
rather than to individual particles. Wherever human modication is
considerable, this is not the natural land surface, which leads to
difculties discussed below. Exclusion of caves etc. is a necessary
simplication to give a single-valued surface as a function of horizontal coordinates, permitting representation by a DEM. Here DEM is
used in the narrower sense, as in Hengl and Reuter (2009): elevation
(altitude) values for a gridded set of points in Cartesian coordinates.
Technological progress has made DEMs an increasingly important
basis for geomorphological research.
3. Dening the land surface on DEMs
Until recently, reliable measurements at 1 to 10 m scale required
eldwork. Early DEMs were coarse (with horizontal resolutions grid
mesh spacings around 100 m) and mainly derived from contours
manually digitised from medium scale maps (Tobler, 1969), e.g. at 1:
50,000 or 1: 25,000 or even small-scale, 1: 250,000. Better quality
was obtained by photogrammetry, e.g. the 10 m DEM used by
Hancock et al. (2006), or at large scales for engineering works or from
photos taken in the eld (Lane, 1998), covering limited areas. It is well
known that mean and standard deviation of slope gradient reduce
rapidly as measured from coarser grids (Evans, 1972: Deng et al.,
2007). Although altitude and regional variables may be insensitive to
scale (Shary et al., 2005), curvatures are affected even more than
gradients: an order-of-magnitude gap remained, between most DEMbased calculations and 1 m scale eld measurements. Having rejected
the fractal model in this context (Evans and McClean, 1995; McClean
and Evans, 2000), it was clear that extrapolation was dangerous. In
particular, gradients and curvatures from the coarse DEMs of the

1970s and 1980s (50 m or 100 m mesh: Evans, 1980) could hardly be
compared with eld-based measurements of slope proles (Young,
1972; Parsons, 1988; Cox, 1990).
Sampling adequacy was also of great concern, together with the
effects of errors in DEMs interpolated from contours (Wise, 2007). For
example, as streams typically do not cooperate by passing through
grid points, the related low points could often be missed, creating
spurious pits upstream. Smoothing does not solve this problem, and
lling the sink by raising levels up to the apparent outlet creates
artefactual plains and falsies data at many grid points. Rather than
raising points over an area, it is better to lower points along a drainage
line, as this is equivalent to displacing them (by less than one grid
mesh) to the position of the presumed nearby channel, and fewer data
points are modied. Pits can be removed by breaching (Martz and
Garbrecht, 1999) or carving (Soille et al., 2003), yet it is dangerous to
apply this automatically as even in uvial topographies real depressions occur, becoming very common not only in karst areas but
also in areas of eluviation of nes, or of deation (Reuter et al., 2009;
Li et al., 2011). A large body of literature grew up on drainage tracing
from DEMs (Gruber and Peckham, 2009), but the possibility remained
of major mis-assignments of drainage direction because of channels
missed between well-spaced grid points. These algorithms could not
match the accuracy of drainage nets plotted directly from air photos.
With the development of two types of active remote sensing,
things are now very different (Nelson et al., 2009; Reuter et al., 2009;
Smith and Pain, 2009). Radar-based products are available at various
resolutions. Airborne Interferometric Synthetic Aperture Radar
(IFSAR) can be own at high altitudes and fast speeds, covering
large areas with few problems from weather. Western Europe and the
conterminous USA are covered by the 5 m resolution NEXTMap IFSAR
DEM. For Britain this was provided in two versions, the initial product
(digital surface model: DSM) showing forests, hedges and buildings,
and a smoothed product (digital terrain model enhanced: DTME)
where these had been ltered out. Livingstone et al. (2008) mapped
drumlins, glacial lineations and other features in much of northern
England, manually interpreting hill-shaded NEXTMap DEMs with two
orthogonal illumination directions, as well as a slope gradient map.
Interestingly, they worked mainly from the initial DSM product
because it retained relevant detail lost in the ltering process. A
number of workers using visual interpretation have preferred to use
the unltered DSM, but it would be difcult to automate landform
recognition without ltering out vegetation and buildings.
Laser scanning can give such dense point clouds that gridding at 1
m represents generalisation. Once we could process last returns as
well as rst returns, the ground and not just the forest canopy could
be mapped accurately from the air, and absolute vertical accuracies
better than 0.3 m could be achieved (Pirotti and Tarolli, 2010). Under
dense forest, however, the bare earth DEM is interpolated from fewer
points and is less accurate than elsewhere because only a small
proportion of returns are from the ground (Norheim et al., 2002).
Although airborne laser scanning (LiDAR: Light Detection and Ranging) is expensive, it is so promising, especially for hazard evaluation,
that coverage of laser-based DEMs with 1 m resolution is rapidly
increasing: for example, for the whole of Switzerland below 2000 m,
and for entire Lnder in Austria. Belgium, the Netherlands and Alberta
are covered at 2 to 5 m resolution. Terrestrial laser scanning can provide even greater detail for individual landforms, and even for coarsegrained sediments (Hodge et al., 2009): in studying overland ow,
(Smith et al., 2010) use a 2 mm resolution DEM. With ne-resolution
LiDAR, many old problems fade, and new ones appear. Error
assessment is changed radically (Fisher and Tate, 2006): technical
errors have become rather small, but errors of denition remain and
loom large.
Laser and radar returns produce data some of which is of
geomorphological interest, and some not. We have to think even
more carefully, exactly what is the land surface? What version of the

I.S. Evans / Geomorphology 137 (2012) 94106

surface is relevant to the problem in hand? We can agree that trees

and bushes should be removed, but what about buildings? Here views
may diverge between those attempting to explain the natural land
surface, and those working on applications such as ood routing and
slope stability. If buildings are solidly built of stone or brick, they will
obstruct water ow up to the point where the walls collapse. Cellars
may extend below the natural ground surface, and the land around a
building will often have been levelled leaving a steeper cut on one or
more sides. This makes it difcult to interpolate through a building
from the surrounding surface, and algorithms to remove buildings
may often leave ghostly artefacts betraying building position.
Working at 1 m resolution, or even 5 m, major road and rail cuts
are clearly represented: removal is difcult and perhaps unreasonable. The anthropogenic nature of much of the modern-day land
surface must be accepted. Rather than being removed, such features
should be identied and delimited as specic landforms, analysed
separately from the remaining surface.
The complexity of desirable ltering algorithms is formidable if we
consider not only buildings and trees, but also vehicles, livestock and
feeding troughs, and derelict parts of these. Producing a LiDAR DEM
compatible with the above concept of land surface seems to require a
full initial remote-sensing identication of all other objects in the data.
One implication of LiDAR is that the amount of data to be processed can be huge: fortunately, the relentless increase in computer
power makes this acceptable. Although the distinction between
point-based and pixel (area-based) raster data is important for
analysis (Strobl, 2008), the generation of 1 m raster grids from LiDAR
point clouds with several returns per m2 does produce some convergence. Another implication of such highly detailed grids is that stream
banks are often visible, and only the tiniest stream channels can be
missed. Pirotti and Tarolli (2010) nd it best to smooth or thin 1 m
DEMs to map small channels and channel heads successfully, from
points with curvature deviations exceeding 2 standard deviations.
This means that gross errors in drainage tracing should not occur. But
representation of the water surface means that results depend on the
water level (stage) at the time of survey. As the true land surface
includes the channel bed (also lake beds) rather than the uctuating
water surface, grid points falling on the surface should be coded as
such: survey of the water bed requires different techniques and is not
available unless bathymetric LiDAR is used (Hilldale and Raff, 2008).
4. Delimiting landforms
Landforms are one type of geomorphological (or geomorphometric) object (Schmidt and Dikau, 1999: MacMillan and Shary,
2009). Other types are linear features: ridge lines, slope lines, course
(valley) lines and break lines, and special points: peaks, passes and
pits. Landforms in this narrower sense are areal objects on a DEM, and
in general they have a third dimension they are volumetric. Mark
and Smith (2004 Table 3.1) dene 25 broad specic landform types,
with many subsets. Some features have multiple identities: all large
volcanoes are mountains, and some volcanoes are islands, but many
islands and mountains are not volcanoes. In geomorphometry and
geomorphological mapping, the more precise term volcano is usually
adopted, rather than the broader terms. Full classications recognise
nested hierarchies of terms: for example, barchan dunes are a type of
transverse dune, a subset of aeolian dunes. Other processes also
produce dunes, which are a mobile type of bedform, and bedforms are
a major type of landform.
Evans (1987) specied nine stages in a specic geomorphometric
study, starting with conceptualisation and operational denition of
the landform type. Further operations can now be performed onscreen or by computer calculation: delimitation, measurement,
derivation of ratios, assessment of frequency distributions, interrelation and mapping. Finally comes the more subjective procedure of
interpretation in relation to genesis and chronology. Concern here is

97

especially with conceptualisation, operational denition and delimitation. As yet, delimitation of specic forms is mainly manual, using
visual identication: automation of the process remains a research
frontier, with acceptable success rates somewhat elusive. Progress is
being made with both supervised and unsupervised classication of
DEMs (Seijmonsbergen et al., in press).
Islands and lakes form special cases that can be delimited precisely
by the water level at a given time, although this does vary a few m even
within a year, and much more over the long term. Nunataks are
likewise limited by the ice surface, which also varies over the years. On
the other hand it could be argued that the true landform of an island or
nunatak extends below the water or ice, for example giving the true
height of Mauna Loa, Hawaii (over the Pacic Ocean oor) as 10,099 m.
The more clearly a landform can be dened, the more likely is it to be
the subject of a morphometric study, i.e. of specic geomorphometry
(Goudie, 1990; Evans, 2010). In glacial geomorphology, this has meant
drumlins, lake basins and cirques, but studies have been extended to
ribbed moraine, megascale glacial lineations, and troughs.
Drumlins and many other depositional or mobile forms have
limited vertical dimension and are, thus, poorly portrayed on standard
contour maps. Fieldwork and air photo interpretation have been
essential in achieving accurate mapping (Rose and Smith, 2008).
Although drumlins are clearly dened, multi-convex landforms,
Smith and Wise (2007) showed that the number identied increased
considerably as satellite image resolution increased (from 30 m to 15
m). Probably the same applies to increasing DEM resolution. Smith
et al. (2006) used visual interpretation of LiDAR data gridded at 2 m
both to check and to improve on eld-based mapping of drumlins.
Differences occurred in both directions, but the LiDAR data permit
subtle hummocks and lineations to be noticed that were missed in the
eld. Bedrock distribution, however, is less clear on the LiDAR images
than in the eld. DEMs from digitised contours on 1: 50,000 and 1:
10,000 maps were inadequate for drumlin delimitation, and satellite
radar data at both 90 m and 25 m horizontal resolution provided no
useful data (Smith et al., 2006 Fig. 4).
Unfortunately the LiDAR data were available for only a small area
and both this and later British studies made extensive use of
NEXTMap airborne radar-based DEMs at 5 m resolution. Fig. 1
compares the delimitations of this set of landforms, showing both
superiority to eld data and some omissions (Clark et al., 2009). While
not as detailed as LiDAR data, NEXTMap has proved a very useful
source for the manual identication of glacial depositional forms and
has shown patterns which had not been recognised in the eld or
from air photos. The identication of drumlins requires high-quality
data, but once recognised they should be relatively easy to delimit
(either manually or automatically).
Smith et al. (2009) have developed a cookie cutter tool for rst
approximations to the volume of convex (or concave) landforms in
relation to a horizontal plane. They used it to calculate the volumes of
drumlins by superimposing digitised outlines onto the 5 m DEM. The
outlines are delimited manually, so the technique is described as
semi-automated.
Visual interpretation of elongate features from a hill-shaded model
is biased by the direction of illumination used (Smith and Clark, 2005;
Smith and Wise, 2007). It is necessary to use multiple directions I
suggest at least four, as does Smith and to combine interpretations
from the different models. This can be cumbersome, and it would be
desirable to automate boundary delimitation by direct processing
of the DEM. A map of slope gradient is equivalent to a vertically
illuminated model (with contrast enhancement), and it eliminates
bias. The outer boundaries of drumlins are essentially concave breaks
in slope, whether they are surrounded by lower areas or contiguous
with other drumlins. Slope gradient steepens away from the summit:
if this prole convexity is followed by a basal concavity, postglacial
redistribution of material by wash or creep is suspected. It is usually
assumed that postglacial modication has been slight, except where

98

I.S. Evans / Geomorphology 137 (2012) 94106

rivers or former meltwater channels have cut into drumlins and left
obviously anomalous steep bluffs. The main complications are where
younger drumlins are superimposed on larger ones that have been
only partially remoulded, and where long streamlined tails merge
into a hillside. Many drumlins contain some bedrock, but Clark et al.
(2009) exclude those which are entirely bedrock and focus on
drumlins formed in deformable materials.
Most bedforms are mobile and should, therefore, be categorised as
neither depositional nor erosional. Fluvial and aeolian bedforms (such
as dunes and bars of various types) have more in common with each
other than either class has with glacial bedforms. They also evolve
faster and are more accessible to study, so that the growth and decay
of individuals can be studied over time, unlike most bedrock erosional
forms. Dunes vary from the easily delimited barchan, where mobile
material is sparse, to the complicated coalescent patterns of akl and
compound dunes.
Volcanic and meteoritic craters are the most easily dened
erosional features, because of sharp convexities often separating
opposing slopes. Yardangs and whalebacks are fairly well delimited.
Another easily delimited form is a lake basin, but if we are interested
in the rock surface underlying its sedimentary ll, expensive geophysical survey is required. Pingos usually have a sharp basal concave
break in slope that clearly delimits them. Volcanoes are rather more
difcult because broad pediments of lava or lahar deposits may merge

gradually with surrounding gentle slopes. Craters, volcanoes, pingos,

dolines and salt domes are unusual in tending to radial symmetry,
whereas most bedforms, plus cirques and slope forms such as landslides, tend to bilateral symmetry about the downstream/downslope
direction. Among karst features, poljes are more irregular and difcult
to delimit than dolines and tower karst (mogotes): likewise alas
depressions in thermokarst (on permafrost) are not so easy to delimit.
Landslide scars are delimited by sharp convexities; landslide deposits,
by concavities which are initially sharp unless the material is so uid
that the lower boundary is unclear. Landslides have clear geomorphometric signatures (Pike, 1988).
Cirques (Fig. 2) are easily dened around the crestal convexities,
which are sharp unless modied by overriding ice. The aim is to
recognise simple, double-concave (plan and prole) features (Fig. 3:
Rasemann et al., 2004). The down-valley limits are no problem for
cirques with clear thresholds (convex breaks in slope below the oor)
but difcult for some valley-head cirques without clear lateral convex
slope changes (buttresses). Sometimes more than one down-valley
limit is feasible not so much a fuzzy boundary (see below) as two
or three alternative boundaries. But in most cases a decision can be
made such that the headwall curves around the oor, and every cirque
(including nested ones) has its own oor and a distinct section of
headwall (Evans and Cox, 1995). The main problems arise from cirque
coalescence around a valley-head (splitters and lumpers may differ
on the number of cirques recognised on the prominence of the
buttress required to recognise two instead of one) and from cirque-incirque forms (each lower cirque must have its own section(s) of
headwall as well as a distinct oor).
Two topographical terms used for steep-sided valleys in northeast
England are dene and gill. Inspection of their use in proper names on
maps of East Durham shows that application is consistent. Denes
start around 100 m altitude and their channels slope at 14 to 30 m
km 1: their active development continues, eroding into both glacial
deposits and Permian dolomite bedrock. Valley-side slopes are
generally steeper than 14 degrees. Similar valleys shorter than 700 m
are known as gills. (Evans, 1999 p. 57). Sharp convexities at the crests
of these valleys delimit them clearly. Gullies representing a change in
process, such as the onset of soil erosion, are similarly delimitable.
Likewise glacial meltwater channels have steep bluffs with upper
convexities and rather at oors.
Returning to What is a mountain?, the philosopher W.V. Quine
proposed an interesting if compound denition. He advocated keeping
the absolute term but resolving its vagueness by arbitrary stipulations:
we may dene a mountain as any regionsuch that
a) the boundary is of uniform altitude.
b) the highest point, or one of them, is at an inclination of at least ten
degrees above every boundary point and twenty degrees above
some, and is at least a thousand feet above them, and
c) the region is part of no other region fullling (a) and (b). (Quine
1981 p. 33).

Fig. 2. Sale Pot, (251E,5430N) in the High Street Range above Hawes Water,
Northern England, is a welldenedvalleyhead cirque with a minor bog and a clear
threshold; it is eroded into mixed Borrowdale Volcanic rocks. Its drainage divide
(dashed), crest (solid line), downvalley limit and oor: headwall boundary (dashed)
are indicated. Contours are shown every 50 m. F: Cirque focus, middle of threshold. The
median axis is drawn from F so that area on left = area on right. It intersects the crest
at 6. Length (of axis) = 590 m; width (normal to axis) = 790 m. Axis aspect = 121;
wall aspect = 085 (steeper wall faces north and east). Altitudes: 1. lowest, 520 m;
2. oor modal, 525 m; 3. max oor 582 m; 4. max crest 772 m; 5. max above 792 m;
6. median crest 752 m. 7. Grid reference, middle of axis: 3443 east, 5123 north. 8. 30 m
fall in 18 m: max gradient = 59; 9. 10 m fall in 195 m: min gradient = 2.9. 10. Wall
height = 770-570 = 200 m (max along any slope line). C: the col here cuts two 10 m
contours, but 20 m this is below the minimum depth requirement (30 m), hence
col = 0. MMM Terminal moraine. planclos = (360-298) + 110 = 172 (measured
over 100 m lengths, the mid-height contour, at 650 m, heads toward 298, turns
clockwise through north, and reaches 110).

The vagueness Quine considered is of three types: of acceptable

altitude, of basal boundary, and of indecision whether two summits
count as two mountains or one (cf. the distinction between Munro's
separate mountains and tops: currently 283 Scottish mountains are
recognised as being separate and over 914.4 m altitude). Separateness may be based on distance as well as altitude change along a
connecting ridge but, following J. R. Corbett, Dawson (1992) has used
a rise of 152.4 m in all directions (rounded to 150 m) to list 1542
separate hill and mountain summits in Great Britain, regardless of
altitude. He includes 205 mountains over 914.4 m altitude in Scotland.
Dawson mentions others who have listed hills or mountains with
summit magnitudes (rise in all directions) variously exceeding 10, 15,

I.S. Evans / Geomorphology 137 (2012) 94106

99

Fig. 3. Denition of cirques and cirque oors in the Iezer Range, Romania. From Cirque denition map of Iezer Mountains.
North is at top. From Mndrescu et al. (2010).

30, 100 and 150 m. This concept of height above the lowest closed
contour within which a peak is the highest point prominence or
summit magnitude is now widely used by mountaineers (http://
www.peaklist.org/WWlists/WorldTop50.html). It permits nested
mountains, contrary to Quine's point (c).
Quine concluded with a theorem:
the boundary of a mountain is the outermost contour line that
lies wholly within ten degrees of steepness from the summit and
partly within twenty.
We may argue with the arbitrary summit magnitude thresholds,
with Quine's arbitrary (10, 20, 1000 feet) thresholds, or with the
need for a contour as the boundary. Macmillan (personal communication 2010) prefers a boundary inclined down-valley. This is relevant
to a binary distinction between mountain and valley. Those starting
from supposedly equivalent terms to mountain in other languages
are very likely to propose different thresholds (Mark and Smith,
2004). But unless our denition includes some such arbitrary criteria,
it is unlikely to be operational that is, to be applicable consistently in
practice. The most fundamental property of the type mountain is

being convex (Mark and Smith, 2004) and large/high: each of these
can be measured quantitatively and can, thus, be the basis of a
denition if international scientists can agree, whatever the variations
in common speech. It must be accepted, however, that mountain is a
vaguer term than cirque, drumlin, crater or landslide, and different
denitions may be useful in different contexts.
At the least, each geomorphological study should state specically
the landform denitions it is attempting to employ. Terms such as
small, short, light or weak are usefully kept relative, not absolute, but
landforms should not be dened in a relative (contextual) way if they
are to be mapped or measured. Thus, operational denitions are
needed for all landforms to which specic morphometry is applied. In
this respect, some studies are more opaque than others: the need for
transparency is axiomatic. Geomorphological mappers have a poor
record in providing operational denitions of the terms in their map
keys.
Currently most landform delimitation is by on-screen digitising of
manually identied boundaries, on maps, satellite images or rectied
air photos (Clark et al., 2009). Much can be done using Google Earth.
Two decades after the pioneer efforts of Tribe (1991), automated
identication of specic types of landform remains difcult (van
Asselen and Seijmonsbergen, 2006). We should be able to do better, at

I.S. Evans / Geomorphology 137 (2012) 94106

5. Fuzzy boundaries and leftover spaces

Even with precise operational denitions, considerable uncertainty remains. Given high-density data such as DEMs, it is realistic to
recognise that some parts are more difcult to classify than others. It
is easy to dene mountain (peaks) as points, and valleys (thalwegs,
not following the sinuosity of channels on oodplains) as lines, but
assigning an area to either of them leaves much room for uncertainty
and disagreement. Fuzzy classication maps the degree of membership to a set of recognised classes (Irvin et al., 1997; Burrough et al.,
2000; MacMillan et al., 2000). For example, Fisher et al. (2004)
classied the surface of the English Lake District in terms of afnity to
six topologically distinct classes of form: peaks, ridges, passes, planes,
channels and pits. Afnity varied with scale: a channel at one scale
might be located on a plane or peak at a broader scale or vice versa.
A single map (instead of six) could be produced by assigning each
point to its modal class, over a range of horizontal scales (from 100 to
3700 m). An entropy map showed the degree of uncertainty in this
assignment.
Deng and Wilson (2008) also measured peakness across a range
of spatial scales. This was based on four criteria: local relief, local mean
slope gradient, relative altitude in a wider area, and (low) number
of local (competing) summits. Each property was related to its
maximum and minimum at that scale, and combined as a weighted
sum between 0 and 1. Various modiable thresholds are involved.
Properties of the class peak at each scale provided a comparator from
which the peakedness of each pixel could be calculated, providing
fuzzy areas of peak entities. These are fuzzy in terms of boundaries,
varying typicality and non-uniform contents. I prefer to keep the term
peak for a point, but these fuzzy areas can provide a further denition
of mountains.
In a glaciated mountain area, Arrell et al. (2007) classied slopes of
varying degrees of divergence, convergence or planarity, and ridges, at
four scales (horizontal resolutions, grid meshes from 50 to 400 m).
Each classication was defuzzied by using the class for which
membership (on a 01 scale) was greatest, except that if the ratio of
the second membership to the largest exceeded 0.6 the point was
identied as an intergrade. They found that ridges and low-gradient
planar areas extreme morphometric classes were persistent, in
that they occupied similar areas at all resolutions. The various types of
slope, on the other hand, varied considerably with resolution. The
classes identied could be regarded as types of elementary form.
The main value of the fuzzy classication concept seems to be the
recognition and measurement of uncertainty in classication of pixels.
If the concept were applied to whole landforms, specic geomorphometry could be applied in the traditional way only once defuzzication was applied. A more advanced application would be to delimit a
core and periphery of each landform, or a series of outlines based on
different membership thresholds, and produce multiple measures
of attributes such as length, height and gradient, leading to a best
estimate (e.g. weighted average) bounded by a range of uncertainty.
That this requires considerably more computation should no longer
be an obstacle. Scatter plots would become plots of crosses showing

uncertainty in x and y (as, for example, in sea level chronologies)

rather than points.
I know of no specic geomorphometry study where the characteristics of the leftover land surface between the delimited landforms
are analysed. This has seemed irrelevant to studies focused on the
identied landforms. It may, however, be relevant to the interpretation of differences between regions: the specic landforms are components of a land system at a broader scale that incorporates the
whole surface. In the case of drumlins, whether they rest on a plain, on
undulating topography, or over a series of transverse valleys, makes a
big difference to their development and morphometry. For context,
therefore, I recommend relating specic geomorphometric studies to
the general geomorphometry of the region and/or of the excluded
land surface.
6. Landform scale-specicity and allometry
Two general concepts of size and shape, applicable to any set of
specic landforms, are of interest. Given a set of measurements of a
type of landform, many workers quote the range in size of most of
the forms, which is rather vague; the overall range, maximum
minimum, is precise but is an unstable and often unrepresentative
statistic. The inter-quartile range is better, but pays no detailed
attention to the tails of the distribution. The trouble with using

Cirque size variables, B.C.

5000

Dimensions (m), logarithmic scale

least for all forms bounded by convex and concave breaks. A

promising combined approach is manual identication of a point or
line within a landform, followed by algorithmic delimitation of a
boundary (Schneider and Klein, 2010). Further approaches might be
adopted from the pattern recognition literature. Automation of
morphometric mapping has been achieved (Hengl and Reuter,
2009), and automation of morphographic mapping should not be
difcult. Mapping of landforms, however, involves morphogenesis,
and it is often necessary to perform eldwork with geophysical,
sedimentological and stratigraphic analyses to be condent of
morphogenesis.

2000

Length

1000

Width

500

Amplitude

200
100

546 cirques: Cayoosh,

Bendor & Shulaps Ranges
-3

-2

-1

Standard normal deviates

b
Dimensions (m), logarithmic scale

100

Cirque size variables, Britain

2000
1000

Width

500

Length

200

Amplitude

100

418 cirques:
Wales & Lake District
-3

-2

-1

Standard normal deviates

Fig. 4. Quantilequantile plots of cirques in (a) British Columbia (Cayoosh, Bendor and
Shulaps Ranges of the southern Coast Mountains), and (b) Wales and the English Lake
District. The x-axis is in standard deviation units. For the method used see Cox (2005).

I.S. Evans / Geomorphology 137 (2012) 94106

symmetrical pairs of other percentiles, such as 9010, is that so many

alternative pairs exist. Thus, I prefer to use the standard deviation, on
a scale which is not too skewed. For most size variables, this means
using a logarithmic scale to reduce positive skewness. (L-moments
calculated using linear combinations of the ordered data reect all
percentiles and provide promising, robust alternatives to standard
deviation and moment-based skewness (Hosking, 1990). They have
often been used in hydrology and climatology but have as yet seen
little application in geomorphology.)

a
length regression
width regression
amplitude regression

length (m)
width (m)
amplitude (m)

5000

2000

1000

500

200

100

546 cirques,
B.C. Coast Mountains
200

500

1000

2000

Size (m)

b
length (m)

length regression

width (m)

width regression
amplitude regression

amplitude (m)

2000

1000

500

200

100

418 cirques,
Wales and Cumbria
200

500

1000

Size (m)
Fig. 5. Allometric plots of cirques in (a) British Columbia (Cayoosh, Bendor and Shulaps
Ranges of the southern Coast Mountains), and (b) Wales and the English Lake District.
Vertical scales are dimensions in metres.

101

For both drumlins and cirques (Evans, 2010 table 5 and p. 145) the
resulting geometric standard deviations of length are around 0.2 on a
log10 scale; this is equivalent to multiplying or dividing by 100.2, i.e. by
1.6. Combined with symmetrical histograms, and linear quantile
quantile plots (Fig. 4), this is good evidence that the distributions are
unimodal and the landforms are scale-specic. The log-normal (logGaussian) model may be accepted as a rst approximation, and
regressions and correlations should be performed on logarithmic
scales. The main differences between the two data sets in Fig. 4 are
that the British Columbian cirques are longer (geometric mean 705 m
cf. 591 m for the British) and higher (geometric mean amplitude 301
m cf. 221 m). Geometric mean widths are 670 m and 674 m respectively. Geometric standard deviations are higher for British Columbia,
but all are between 0.16 and 0.21 (log10 scale).
Evidence from the literature, reviewed for example in Evans
(2003), suggests that similar considerations apply to other glacial and
uvial bedforms, to sinkholes (dolines), karren, karst towers, tors,
impact craters, pingos, volcanoes and to some tectonic forms: all are
scale-specic, regionally if not globally. The magnitudefrequency
distributions of landslides, which cover a greater range of scales than
cirques, have of late generated some controversy. They follow a power
law (a negative Pareto distribution), but only over a limited range of
scales. With due allowance for the censoring of distributions
distortion by the incompleteness of detection of smaller features or
the infrequency of very large features I conclude that landslides too
are scale-specic, especially for particular types or for clusters
produced by single events (Evans, 2010). The lower size limit reects
a threshold critical mass: the upper size limit may reect the frame,
the available slope relief.
A further type of scale-specicity is where breaks in slope occur in
plots of frequency against size, as in some landslide distributions
(Brardinoni and Church, 2004), or of one dimension against another,
for example the depth and diameter of impact craters (Pike, 1980). In
these cases, scaling is combined with scale-specicity. More generally,
scaling is observed within the single order of linear magnitude (two
orders of areal magnitude) embraced by drumlins or cirques. Fig. 5
shows how different size variables scale with a combined, overall
measure of size (the cube root of length width vertical amplitude).
Again the greater lengths and amplitudes in British Columbia are
noticeable, but the gradients of the three logarithmic regressions are
very similar in the two study areas. The vertical amplitude of cirques
increases with overall size more slowly than do length and width.
With 95% condence limits, the respective exponents are 1.10 0.04
and 1.14 0.04 for length, 1.03 0.05 and 1.03 0.07 for width, and
0.88 0.05 and 0.83 0.07 for amplitude, for British Columbia and
Wales plus Cumbria respectively. The condence intervals on amplitude exponents are far from overlapping the others, showing that rate
of change of amplitude with size is lower than that of length and
width, at a high signicance level.
Cirques exhibit static allometry (Evans, 2006b; 2010): the shape of
cirques varies with size. Unfortunately, it is not possible to observe
change over the long periods of time involved in cirque formation. For
rapidly changing bedforms, such as aeolian and uvial dunes, frequently repeated eld survey might test allometric growth directly.
With modern observation techniques dynamic allometry true
allometric growth can be established for mobile bedforms in the
laboratory. Nevertheless the static allometry of scale-specic landforms shows that scaling is compatible with scale-specicity. This
implies that the limits to observed scaling should always be stated;
the existence of limits generates further hypotheses.
7. Hillslopes in uvial basins: what are the landforms?
Scaling over broader ranges is observed in drainage networks.
These have a long history of manual analysis by specic morphometric
methods (Chorley, 1969). Much of this is based on lines and networks

102

I.S. Evans / Geomorphology 137 (2012) 94106

rather than areas. Channels, oodplains, landslides and fans are easily
delimited and have been studied by specic geomorphometry, but
hillslopes have more often been treated differently. Although
hillslopes or hillsides are generally recognised as landforms, they
have rarely been analysed as areal patches. Rather, they have been
studied as proles (Pitty, 1971; Young, 1972). This may be because in
well-dissected terrain hillslopes cover almost all the landscape; or
because although it is easy to delimit upper and lower boundaries (at
ridges or crests and channels or oodplain margins), lateral boundaries are generally indistinct. Nevertheless, proles can only take us so
far in analysing the 3-dimensional surface and further methods are
needed.
MacMillan and Shary (2009) illustrated ve different classications of form elements (or of local surface shape) based essentially on
plan and prole curvature: for example convex, straight or concave in
plan, and in prole. They pointed out a disadvantage in schemes that
recognise straight in plan: the results depend on whether plan
(contour) curvature or tangential curvature is used. Any of the ve
classications can be applied to any point on the surface, and mapped
automatically: they are properties of the land surface rather than
classes of landform. By tting generalised surfaces, they can be
applied to elementary forms in the sense of Minr and Evans (2008).
Also many landforms can be forced into one or other of these curvature categories, but other properties such as elongation, gradient
and position may be more important in dening landforms. Most
geomorphologists use variables and form classes related to the gravity
eld, but Shary (1995) and Shary et al. (2002, 2005) also recognise a
set of eld-invariant morphometric variables such as unsphericity,
mean curvature, and total Gaussian curvature, and forms such as Cdepressions, dimples in the surface that are open and, thus, do not
accumulate water.
Starting from DEMs, surface-specic points (peaks, passes and
pits) and lines (ridges and channels) can also be dened automatically
at a given scale. Break lines and inections, however, are more useful
in delimiting basic units. These small patches of near-uniform
morphometry are termed elementary forms by Minr and Evans
(2008): see the application by Mentlk et al. (2010). They are basic
geomorphological objects; they cover the whole land surface and each
is united by homogeneity in altitude, slope, curvature or change in
curvature, and bounded by break lines or inections in one or more of
these local surface derivatives. The positions of elementary forms
within the hillslope and drainage basin may then be used to specify
the broader spatial structure, and used as the basis for classication.
Drgu and Blaschke (2006) rst dene forms from homogeneity in
elevation, prole curvature, plan curvature and slope gradient, and
then classify these in terms of slope position. Given ne-resolution
data, such as DEMs with 1 m grid mesh, classication of the whole
surface directly from pixels or grid points is inefcient and potentially
misleading: areal geomorphological objects should be generated rst
(Drgu and Blaschke, 2006; Strobl, 2008).
Denition of elementary forms has as yet been subjective, part of
geomorphological mapping (Mentlk et al., 2010). Theory suggests that
more objective, repeatable denition should be approached from two
directions: the recognition of breaks in slope and curvature to provide
boundaries, and the measurement of internal uniformity. A basic
problem, still to be solved, is the trade-off between the sharpness of a
break, the degree of a change, and the lateral continuity, all of which
make it more useful as a landform or elementary form boundary. Clear
boundaries rarely close completely around a landform so, as in the case
of cirques, compromises must often be made in closing them.
The landscape position of elementary forms can be described as
upper, middle and lower slope or footslope, and terms such as crest/
interuve, buttress/nose, hollow/open depression, hillock and ridge
can be applied (Speight, 1990). Types of elementary form that take
position into account may be termed land elements, after Schmidt and
Hewitt (2004) and Schmidt and Andrew (2005). MacMillan and Shary

(2009) have applied such classication to large areas, for ecological

interpretation, and Drgu and Blaschke (2008) have satisfactorily
classied coarse (90 m) radar data. Association of land elements in a
toposequence leads on to the landscape scale of analysis. Skidmore
(1990) quantied slope position in a DEM by rst dening a ridge
network and a stream network. Relative position is (distance to
nearest valley cell)/(sum of distances to nearest valley cell and
nearest ridge cell). The critical phase here is agreeing on the limits of
the two networks. MacMillan et al. (2009) analysed vertical as well as
horizontal relative position.
Whereas Minr and Evans (2008) emphasise homogeneity in local
properties before considering position, MacMillan et al. (2000; 2004)
and MacMillan and Shary (2009 gs. 10 and 12) recommend dening
landform elements on the basis of slope position as well as surface
derivatives. They start by delimiting local catchments of downslope
and notional upslope ow and intersecting these to dene hill sheds.
These are landforms (functional regions) rather than elementary
forms (formal, homogeneous regions), and are likely to contain
several of the latter. See also the valley-side basins delimited by
Rasemann et al. (2004). Hill sheds could form the basis of repeated
landform patterns (land systems). The use of drainage tracing is also
at the basis of the TAPES-C approach to uvial landscapes outlined and
applied in Wilson and Gallant (2000).
Such a mosaic map, dividing an area exhaustively and uniquely
into sets of qualitatively different patches, can be analysed quantitatively in terms of proximity or contiguity of these sets. Considering the
boundaries between, say, type i and type j, the number of boundary
segments nij can be counted to give a matrix of proximities. This
can be the evidence for a toposequence in three dimensions, downvalley as well as down-slope. More precisely, the length of common
boundary lij can be measured (in a GIS, not manually). It may be
necessary to standardise entries in this matrix by expressing each as a
proportion of the total length of boundaries of i and j combined.
Finally, the characteristics of each boundary segment can be measured
and summarised: the average altitude and the average change in
gradient at each boundary segment, and the difference in average
altitude or other properties between the adjacent patches, can be
assessed.
It would be interesting to see such analysis applied in geomorphology and landscape ecology. In structural geomorphology, Minar et al.
(2011) have measured the change in average altitude across the
boundaries of geomorphometric regions, but few other examples exist.
In summary, classicatory (atomistic) approaches to well-dissected
uvial topography still pose many challenges (Cox, 1978). Until they are
fully solved, the application of prole analyses and of general
geomorphometric techniques to drainage basins remains a safer option.
But uvial topography is not the most difcult terrain for the recognition
of landforms and the application of specic geomorphometry.
8. Plains and difcult terrains
Further topographies exist, not discussed above, where specic
morphometry encounters difculties. One is very extensive plains, as
in savanna regions such as central Chad, where no clear breaks or
changes in the topography occur for tens or hundreds of kilometres.
Also on extensive uvial plains, such as the Indo-Gangetic, the only
interruptions are river channels. Past channels may be mapped, but
their surface expression is often more subtle than anthropogenic
features on the surface.
Rather than recognising very extensive elementary forms, most of
us would revert to general geomorphometry in such situations. Unless
channel beds are included, these plains have very low standard
deviations of altitude and of slope gradient, very low mean slope
gradient and, thus, near-zero prole curvature. Plan curvatures, however, can be extreme witness the intricate contours on oodplains.
It may be useful to recognise regional slopes, governing the general

I.S. Evans / Geomorphology 137 (2012) 94106

103

Fig. 6. A difcult terrain for specic geomorphometry: view northwest from Jebel Musa, Sinai, Egypt.

direction of ow. This would certainly be the case on submarine fans

bordering abyssal plains, as it is for rather steeper subaerial fans.
Etched or scoured, largely rocky surfaces (such as southern Sinai,
Fig. 6) are a different problem. Further examples are the glacially
scoured Shield areas of Canada and Fennoscandia, the etched karst of
Guangxi in south China, the inselbergs of southwest Jordan, and
numerous other granitic terrains. Considering maps such as those in
Thomas (1994), it would be possible to measure the shape, size
and mutual relations of each outcrop, bounded by plain or by major
joints. Alternatively, it would be less work to perform a general geomorphometric analysis of the whole region. Comparison of the two
approaches would be interesting.

density) or relate attributes to stream order, but application of

general geomorphometry implies analysis of distributed attributes
and interrelations.
3. Mountain ranges or other landforms. Mountain ranges are in many
ways the dual of drainage basins, and are delimited by valleys and
low passes (that is, they are more extensive than mountains discussed above). They are natural units for analyses where variation
with altitude is important, as for glacial features (Evans, 2006a).
Mountain ranges are in themselves broad-scale landforms, and
general geomorphometry can be applied also to the variability of
altitude, slope, and curvature within any denable landforms, or
landscapes with repeated landform patterns (land systems).

9. Study areas for general geomorphometry

The land surface can be divided exhaustively and uniquely into

map sheets (approximate rectangles), into drainage basins (bounded
by divides) or into mountain ranges (hills, bounded by thalwegs and
passes). Unlike map sheets, basins and ranges cannot in general have
equal areas, but each can claim to be natural divisions for some
purposes, and they can be dened algorithmically. If a large area were
divided into approximately equal numbers of each of these types, the
summary statistics should not be expected to be similar. Each type of
division can be the basis for quantitative characterization of topography, but basins and ranges are more likely to lead to interesting
comparisons of process relevance.

Given these difculties with specic geomorphometry, general

geomorphometry should be more widely used. It is, indeed, generally
applicable. Also Studies of process are much more easily harmonised with local or global representations of the Earth's surface in
terms of elds of elevations, rather than with quantitative or
qualitative representations of shapes or forms. (Mark and Smith,
2004 p. 81). Early studies (Evans, 1972) were restricted to local
properties (altitude, slope and curvature distributions and interrelations), but once drainage tracing and positional variables could be
computed for large data sets they became a major part of general
geomorphometry (Dietrich et al., 2003; Wilson and Gallant, 2000).
General geomorphometry has an inherent diversity based on the
way in which study areas or cases are dened. I suggest an essentially
three-fold division:
1. Arbitrary areas such as map sheets, rectangles or circles. This is
easiest, as it is the way data are provided, and convenient in that a
series of cases of the same areal extent may be compared. But
boundary effects are a problem when drainage tracing and regional
variables are calculated.
2. Drainage basins. These are most evidently appropriate in uvial
landscapes (Chorley, 1969), but can be useful elsewhere. Many
studies use overall attributes of drainage basins (shape, drainage

10. Geomorphological mapping in a GIS environment

To most Europeans, geomorphological maps mean comprehensive multi-colour, multi-layer maps covering morphography, morphometry, genesis, current processes, materials (lithology, structure)
and chronology, plus a base map, such as the two German series
(Barsch and Liedtke, 1980). Most North American and British
scientists, on the other hand, regard comprehensive geomorphological maps as too complex and difcult to read especially if the
legend covers more space than any map sheet. They see major
problems in integrating so much information on one map, overloading the power of the human visualisation system. They prefer
simpler and clearer maps focused on a group of related landforms

104

I.S. Evans / Geomorphology 137 (2012) 94106

(Evans, 1990), even if this means ignoring other aspects of the land
surface. In other words they are more specialised and more in tune
with specic geomorphometry, but at risk of missing interactions
between different geomorphic systems. Some convergence has
occurred, as the legends of comprehensive coloured geomorphological maps have tended to simplify some components, for example in
Dutch (de Graaff et al., 1987) and Italian examples (Pasuto et al.,
2005) which have little chronological or morphometric information. A
Greek view of traditional geomorphological mapping, abundantly
illustrated and with many proposed symbols, is given by Pavlopoulos
et al. (2009). Further simplication may be required to incorporate
maps into GIS (Gustavsson et al., 2006, 2008).
The rationale of putting everything on one map was to permit
visual interrelation of the different components. This was generally
unsuccessful in correlating spatial patterns because of the nonrelevant information interfered; comprehensive maps permitted data
mining by reading off the various attributes of a particular point,
rather than giving a synoptic view of spatial patterns. A successful
multi-layered GIS (Minr et al., 2005; Gustavsson et al., 2008) avoids
these problems by permitting any one layer to be viewed with any
other, and coded subsets of any layer to be used. Together with the use
of simpler, clearer, more specialised maps, this has revived interest
in geomorphological mapping (Paron and Smith, 2008; Pavlopoulos
et al., 2009; Smith et al., in press). Modern computing has also provided many new ways, including animation, of visualising complex
data (Dykes et al., 2005).
The general-purpose geomorphological map may be dead, but long
live geomorphological maps! With the abundance of data now
available, and the range of visualisation techniques, geomorphological
mapping is more important than ever before.
11. Discussion and conclusions
I have attempted to relate geomorphometry to geomorphological
mapping through their common dependence on dening and delimiting landforms and elementary forms. This leads to fuller development of concepts of specic and general geomorphometry. As yet,
both have rarely been applied together and compared in a specic
landscape. Specic geomorphometry is clearly more applicable to
some landscapes than to others, and the number of studies published
reects these differences. Landscapes of bedforms and of distinct
erosional forms lend themselves to specic geomorphometry of areal
forms, whereas analysis of uvial landscapes is more often linearbased: both, of course, should have a vertical dimension. Point-based
analysis seems less applicable, because most features of interest (most
geomorphological objects except for peaks, passes and pits) have
linear or areal extent. (Point pattern analysis has been applied for
example to drumlins, but it is problematic because each drumlin
covers an area that is large relative to the separation of drumlin
centroids.)
Specic morphometry can lead on to various types of generalisation, including scale-specicity and allometry. We are much more
likely to dene types of feature as landforms if they have limited and
characteristic size ranges. Scale specicity relates either to process
thresholds or to the scale of controlling frameworks (e.g. relief of a
whole valley-side, for mass movements) (Evans, 2003; 2010).
Cox (1978) objected to atomistic approaches to subdivision of the
land surface. Cox makes the case that extracting patches from a
continuous surface is unphysical in that the whole surface forms
together (Cox, 2007, and personal communication, 2010), whether
subaerial, subaqueous, supraglacial or subglacial. Diffusive processes
tend to atten the surface, limiting relief and slope gradients and
blurring boundaries. The demand for subdivision of the surface into
manageable objects, however, continues and even grows for
example among soil scientists (MacMillan et al., 2004; Deng, 2007;
and chapters in Hengl and Reuter, 2009). A psychological need to

subdivide seems to exist: also processes such as undercutting,

incision and faulting produce numerous breaks in slope that provide
useful, non-arbitrary boundaries. In geomorphometry and in geomorphological mapping, both general and specic (i.e. both continuous eld and object-based) approaches are needed, and are
complementary.
One facet that has been missing from most quantitative work is the
spatial pattern of areas (rather than points or channel networks).
Scope exists to develop mosaic analysis for patch maps where the
whole area is allocated to one form or another; contiguities are of
genetic signicance, and relative positions affect current processes.
Some ideas are discussed in Fortin and Dale (2005, ch. 4), in an
ecological context. Further development is needed for patches such as
drumlins and cirques which do not occupy the whole surface:
representation as points is problematic, but Fortin and Dale (2005 p.
6475) discuss denitions of neighbour and the use of Minimum
spanning trees. The importance of land systems implies that we
should relate landforms to context (MacMillan et al., 2004; Deng,
2007), to position within the broader system, but this has rarely been
quantied. Studies of systems are now well entrenched in process
geomorphology: the components produced by geomorphological
mapping and analysed by specic geomorphometry should, accordingly, be reassembled and interrelated quantitatively. Object delimitation should be followed by contiguity and contextual (positional)
analysis and system synthesis.
The land surface of Earth is complex because of the range of
processes that have fashioned it from a variety of materials, and to the
way these processes have changed over time, both interacting and
alternating. Even if the view of the surface is simplied to a singlevalued function of latitude and longitude (no pipes, caves, overhangs
or vertical slopes), and human modications are excluded, the land
surface cannot be represented accurately by any mathematical model
with a small number of parameters (Evans and McClean, 1995;
McClean and Evans, 2000). Such models (e.g. fractal or spectral; also
Fourier series and other families of polynomials) have uses, but it is
dangerous to regard them as realistic, or even as capturing the essence
of a real land surface. The science of geomorphology has devoted
much effort to classifying land surfaces and to recognising specic
surface features (landforms and elementary forms) and measuring
relationships between surface attributes. Specic global and regional
scales are recognised in the form of the land surface. We need to apply
both specic and general geomorphometry, and to supplement
multiple geomorphological maps with numerous geomorphometric
maps, interrelating them in a GIS environment.
Apart from demonstrating the scale-specicity and static allometry of many landforms, this paper has reviewed semantic and
conceptual frameworks for mapping and analysing landforms.
Situations where specic or general geomorphometry are appropriate
have been discussed. The importance of ne resolution DEMs,
especially LiDAR, for this enterprise has been emphasised, and a
number of proposals for further work have been made. Some parts of
the land surface belong to more than one landform, and there are
leftover areas difcult to allocate. Classication, therefore, need not
be exclusive, or exhaustive: nesting and superposition occur, and
hierarchies are common.
Algorithms are needed that produce results that are consistent,
and very accurate in approximating geomorphologists' views of
landforms. We need to move from manual delimitation, whether onscreen or on printed images, to automated recognition and delimitation of landforms and elementary forms from DEMs. (Measurement
of delimited forms is a more easily automated task.) Local properties
such as altitude, slope and curvature are fundamental, but many
forms need to be related to the ow network as position and context
are important for their classication or recognition as land elements
or types of landform. Extensive plains and highly irregular topographies pose special challenges.

I.S. Evans / Geomorphology 137 (2012) 94106

In conclusion, a quotation from Chorley (1969, pp. 9697) has

proved highly prescient:
In the past, morphometric analysis from maps has been a rather
tedious and time-consuming task, . programming promises
to release the masses of data locked up in topographic maps and
will obviously allow much more extensive sampling and generalisation of morphometric properties. Before too long these
methods will be applied directly to the output from aircraft and
satellite scanning equipment, obviating the necessity for the
actual compilation of many maps. The Holy Grail of full
automation remains elusive, but progress has been made.

Acknowledgements
I have drawn on useful discussions with Jozef Minr, Pavel Mentlk
and Lucian Drgu, and with the participants of the excursion Back to
reality Reconciling geomorphometry and geomorphology in the
eld (led by Oliver Korup) from the Geomorphometry 2009
Conference at Zrich, Switzerland. Nick Cox, Nick Odoni, Mike
Smith, Bob MacMillan and an anonymous reviewer generously
provided constructive comments on the submission, and Nick Cox
provided the Dawson, Jeffreys and Quine quotations.
References
Ahnert, F., 1998. Introduction to Geomorphology. Arnold, London. 352 pp.
Arrell, K.E., Fisher, P., Tate, N., Bastin, L., 2007. A fuzzy c-means classication of elevation
derivatives to extract the morphometric classication of landforms in Snowdonia,
Wales. Computers and Geosciences 33, 13661381.
Barsch, D., Liedtke, H., 1980. Principles, scientic value and practical applicability of the
geomorphological map of the Federal Republic of Germany at the scale of 1:25,000
(GMK 25) and 1:100,000 (GMK 100). Zeitschrift fr Geomorphologie N.F.
SupplementB and 36, 296313.
Brardinoni, F., Church, M., 2004. Representing the landslide magnitudefrequency
relation: Capilano River Basin, British Columbia. Earth Surface Processes and
Landforms 29, 115124 doi:10.1002/esp. 1029.
Burrough, P.A., van Gaans, P., MacMillan, R., 2000. High resolution landform
classication using fuzzy k-means. Fuzzy Sets and Systems 113, 3752.
Butzer, K.W., 1976. Geomorphology from the Earth. Harper and Row, New York. 463 pp.
Chorley, R.J., 1969. The drainage basin as the fundamental geomorphic unit. In: Chorley,
R.J. (Ed.), Water, Earth and Man. Methuen, London, pp. 7799. [Also as:
Introduction to Fluvial Processes, pp. 3052.].
Chorley, R.J., Dunn, A.J., Beckinsale, R.P., 1964. The History of the Study of Landforms,
v.1. Methuen, London. 678 pp.
Chorley, R.J., Schumm, S.A., Sugden, D.E., 1984. Geomorphology. Methuen, London. 605 pp.
Clark, C.D., Hughes, A.L.C., Greenwood, S.L., Spagnolo, M., Ng, F.S.L., 2009. Size and shape
characteristics of drumlins, derived from a large sample, and associated scaling
laws. Quaternary Science Reviews 28, 677692.
Cox, N.J., 1978. Hillslope prole analysis (comment). Area 10, 131133.
Cox, N.J., 1990. Hillslope proles, In: Goudie, A.S. (Ed.), Geomorphological Techniques,
2nd edition. Unwin Hyman, London, pp. 9296.
Cox, N.J., 2005. The protean quantile plot. Stata Journal 5, 442460.
Cox, N.J., 2007. Kinds and problems of geomorphological explanation. Geomorphology
88 (1), 4656.
Dawson, A., 1992. The Relative Hills of Britain. Cicerone Press, Milnthorpe, Cumbria. 256
pp. ISBN 1-85284-068-4.
de Graaff, L.W.S., De Jong, M.G.G., Rupke, J., Verhofstad, J., 1987. A geomorphological
mapping system at scale 1:10,000 for mountainous areas. Zeitschrift fr
Geomorphologie N.F. 13 (2), 229242.
Deng, Y., 2007. New trends in digital terrain analysis: landform denition, representation and classication. Progress in Physical Geography 31 (4), 405419.
Deng, Y., Wilson, J.P., 2008. Multi-scale and multi-criteria mapping of mountain peaks
as fuzzy entities. International Journal of Geographical Information Science 22 (2),
205218.
Deng, Y., Wilson, J.P., Bauer, B.O., 2007. DEM resolution dependencies of terrain
attributes across a landscape. International Journal of Geographical Information
Science 21 (2), 187213.
Dietrich, W.E., Bellugi, D.G., Sklar, L.S., Stock, J.D., Heimsath, A.M., Roering, J.J., 2003.
Geomorphic transport laws for predicting landscape form and dynamics. In:
Wilcock, P.R., Iverson, R.M. (Eds.), Prediction in Geomorphology. : Geophysical
Monograph, 135. American Geophysical Union, Washington DC, pp. 103132.
Drgu, L.D., Blaschke, T., 2006. Automated classication of landform elements using
object-based image analysis. Geomorphology 81, 330344.
Drgu, L.D., Blaschke, T., 2008. Terrain segmentation and classication using SRTM
data. In: Zhou, Q., Lees, B., Tang, G.A. (Eds.), Advances in Digital Terrain Analysis.

105

Series Lecture Notes in Geoinformation and Cartography. Springer, Berlin, pp.

141158.
Dykes, J., MacEachren, A.M., Kraak, M.J. (Eds.), 2005. Exploring Geovisualization.
Elsevier, Amsterdam. 730 pp.
Evans, I.S., 1972. General geomorphometry, derivatives of altitude, and descriptive
statistics. In: Chorley, R.J. (Ed.), Spatial Analysis in Geomorphology. Methuen,
London, pp. 1790.
Evans, I.S., 1980. An integrated system of terrain analysis and slope mapping. Zeitschrift
fr Geomorphologie N.F. Supplementband 36, 274295.
Evans, I.S., 1987. The morphometry of specic landforms. In: Gardiner, V. (Ed.),
International Geomorphology 1986 Part II. John Wiley, Chichester, pp. 105124.
Evans, I.S., 1990. Cartographic techniques in geomorphology, In: Goudie, A.S. (Ed.),
Geomorphological Techniques, 2nd edition. Unwin Hyman, London, pp. 97108.
Evans, I.S., 1999. Castle Eden Dene and Blunts Dene. In: Bridgland, D.R., Horton, B.P.,
Innes, J.B. (Eds.), The Quaternary of North-east England: Field Guide. Quaternary
Research Association, London, pp. 5764. ISSN 0261 3611.
Evans, I.S., 2003. Scale-specic landforms and aspects of the land surface. In: Evans, I.S.,
Dikau, R., Tokunaga, E., Ohmori, H., Hirano, M. (Eds.), Concepts and Modelling in
Geomorphology: International Perspectives. Terrapub, Tokyo, pp. 6184.
Evans, I.S., 2006a. Glacier distribution in the Alps: statistical modelling of altitude and
aspect. Geograska Annaler 88 A (2), 115133.
Evans, I.S., 2006b. Allometric development of glacial cirque form: geological, relief and
regional effects on the cirques of Wales. Geomorphology 80 (34), 245266.
Evans, I.S., 2010. Allometry, scaling and scale-specicity of cirques, landslides and
other landforms. Transactions of the Japanese Geomorphological Union, 31 (2),
133153.
Evans, I.S., Cox, N.J., 1995. The form of glacial cirques in the English Lake District,
Cumbria. Zeitschrift fr Geomorphologie N.F. 39, 175202.
Evans, I.S., McClean, C.J., 1995. The land surface is not unifractal: variograms, cirque
scale and allometry. Zeitschrift fr Geomorphologie N.F. SupplementBand 101,
127147.
Fisher, P.F., Tate, N.J., 2006. Causes and consequences of error in digital elevation
models. Progress in Physical Geography 30 (4), 467489.
Fisher, P., Wood, J., Cheng, T., 2004. Where is Helvellyn? Fuzziness of multi-scale
landscape morphometry. Transactions of the Institute of British Geographers 29
(1), 106128.
Fortin, M.-J., Dale, M., 2005. Spatial Analysis: a Guide for Ecologists. 365 pp Cambridge U.P.
Gerrard, A.J.W., Robinson, D.A., 1971. Variability in slope measurements. Transactions
of the Institute of British Geographers 54, 4554.
Goudie, A.S. (Ed.), 1990. Geomorphological Techniques. 2nd edition. Unwin Hyman,
London. 570 pp.
Gruber, S., Peckham, S., 2009. Land-surface parameters and objects in hydrology. Ch.7 in
Hengl and Reuter, 2009 (q.v.) , pp. 171194.
Gustavsson, M., Kolstrup, E., Seijmonsbergen, A.C., 2006. A new symbol- and GIS-based
detailed geomorphological mapping system: renewal of a scientic discipline for
understanding landscape development. Geomorphology 77, 90111.
Gustavsson, M., Seijmonsbergen, A.C., Kolstrup, E., 2008. Structure and contents of a
new geomorphological GIS database linked to a geomorphological map with an
example from Liden, central Sweden. Geomorphology 95, 335349.
Hancock, G.R., Martinez, C., Evans, K.G., Moliere, D.R., 2006. A comparison of SRTM and
high-resolution digital elevation models and their use in catchment geomorphology and hydrology: Australian examples. Earth Surface Processes and Landforms 31
(11), 13941412.
Hengl, T., Reuter, H. (Eds.), 2009. Geomorphometry: Concepts, Software, Applications.
Developments in Soil Science, 33. Elsevier, Amsterdam. 765 pp 13: 978-0-12- 1124
374345-9.
Hilldale, R.C., Raff, D., 2008. Assessing the ability of airborne LiDAR to map river
bathymetry. Earth Surface Processes and Landforms 33, 773783.
Hodge, R., Brasington, J., Richards, K.S., 2009: In situ characterization of grain-scale
uvial morphology using terrestrial laser scanning. Earth Surface Processes and
Landforms 34, 954968.
Hosking, J.R.M., 1990. L-moments: analysis and estimation of distributions using linear
combinations of order statistics. Journal of the Royal Statistical Society, Series B 52,
105124, http://www.jstor.org/stable/2345653.
Irvin, B.J., Ventura, S.J., Slater, B.K., 1997. Fuzzy and isodata classication of landform
elements from digital terrain data in Pleasant Valley, Wisconsin. Geoderma 77,
137154.
Jeffreys, H., 1961. Theory of Probability, 3rd. edition. Clarendon Press, Oxford. 459 pp.
Lane, S.N., 1998. The use of digital terrain modelling in the understanding of dynamic river
channel systems. Ch. 14 In: Lane, S.N., Richards, K.S., Chandler, J.H. (Eds.), Landform
Monitoring, Modelling and Analysis. John Wiley, Chichester, pp. 311342.
Li, S., MacMillan, R.A., Lobb, D.A., McConkey, B.G., Fraser, W.R., 2011. Lidar DEM error
analyses and topographic depression identication in a hummocky landscape in
the Prairie region of Canada. Geomorphology 129, 263275.
Livingstone, S.J., O'Cofaigh, C., Evans, D.J.A., 2008. Glacial geomorphology of the central
sector of the last BritishIrish Ice Sheet. Journal of Maps 2008, 358377, http://
www.journalofmaps.com 10.4113/jom.2008.1032.
Lobeck, A.K., 1939. Geomorphology. McGraw-Hill, New York. 731 pp.
MacMillan, R.A., Shary, P., 2009. Landforms and landform elements in geomorphometry. Ch.9 in Hengl and Reuter, 2009 (q.v.) , pp. 227254.
MacMillan, R.A., Pettapiece, W.W., Nolan, S.C., Goddard, T.W., 2000. A generic procedure
for automatically segmenting landforms into landform elements using DEMs,
heuristic rules and fuzzy logic. Fuzzy Sets and Systems 113, 81109.
MacMillan, R.A., Jones, R.K., McNabb, D.H., 2004. Dening a hierarchy of spatial entities
for environmental analysis and modeling using digital elevation models (DEMs).
Computers, Environment and Urban Systems 28 (3), 175200.

106

I.S. Evans / Geomorphology 137 (2012) 94106

MacMillan, R.A., Torregrosa, A., Moon, D., Coup, R., Philips, N., 2009. Automated
predictive mapping of ecological entities. Ch. 24 in Hengl and Reuter, 2009 (q.v.) ,
pp. 551578.
Mark, D.M., Smith, B., 2004. A science of topography: from qualitative ontology to
digital representations. In: Bishop, M.P., Shroder, J.F. (Eds.), Geographic Information Science and Mountain Geomorphology. Springer, Berlin, pp. 7597.
Martz, L., Garbrecht, J., 1999. An outlet breaching algorithm for the treatment of closed
depressions in a raster DEM. Computers and Geosciences 25 (7), 835844.
McClean, C.J., Evans, I.S., 2000. Apparent fractal dimensions from continental scale
digital elevation models using variogram methods. Transactions in GIS 4(4), 361
378.
Mentlk, P., Minr, J., Bzov, E., Lis, L., Tbok, P., Stacke, V., 2010. Glaciation in the
surroundings of Prilsk Lake (Bohemian Forest, Czech Republic). Geomorphology
117, 181194.
Minr, J., Bielik, M., Kov, M., Plaienka, D., Barka, I., Stankoviansky, M., Zeyen, H., 2011.
New morphostructural subdivision of the Western Carpathians: an approach
integrating geodynamics into targeted morphometric analysis. Tectonophysics
502, 158174.
Minr, J., Evans, I.S., 2008. Elementary forms for land surface segmentation: the
theoretical basis of terrain analysis and geomorphological mapping. Geomorphology 95 (34), 236259. doi:10.1016/j.geomorph.2007.06.003.
Minr, J., Mentlk, P., Jedlika, K., Barka, I., 2005. Geomorphological information
system idea and options of practical implementation. Geograck asopis 58,
247266.
Mndrescu, M., Evans, I.S., Cox, N.J., 2010. Climatic implications of cirque distribution in
the Romanian Carpathians: palaeowind directions during glacial periods. Journal of
Quaternary Science 25 (6), 875888 doi:10.1002/jqs.1363 on-line February 2010.
Nelson, A., Reuter, H.I., Gessler, P., 2009. DEM Production Methods and Sources. Ch.3 in
Hengl and Reuter, 2009 (q.v.) , pp. 6585.
Norheim, R.A., Queija, V.R., Haugerud, R.A., 2002. Comparison of LIDAR and INSAR DEMs
with dense ground control. Proceedings, 2002 ESRI Conferencegis.esri.com.
Paron, P., Smith, M.J., 2008. Editorial: applied geomorphological mapping. Journal of
Maps 2008, 197200, http://www.journalofmaps.com 10.4113/jom.2008.1048.
Parsons, A.J., 1988. Hillslope Form. Routledge, London. 212 pp.
Pasuto, A., Soldati, M., Siorpaes, C., 2005. Geomorphological Map of the Surroundings of
Cortina d'Ampezzo (Dolomites, Italy): Scale 1: 20,000. Universit di Modena e
Reggio Emilia, Modena, Dipartimento di Scienze della Terra.
Pavlopoulos, K., Evelpidou, N, Vassilopoulos, A. 2009. Mapping geomorphological
environments. Springer, Berlin. 235 pp.
Pike, R.J., 1980. Control of crater morphology by gravity and target type: Mars, Earth,
Moon. Proceedings, Lunar & Planetary Science Conference, 11th, pp. 21592189.
Pike, R.J., 1988. The geometric signature: quantifying landslide terrain types from
digital elevation models. Mathematical Geology 20, 491511.
Pike, R.J., 2000. Geomorphometry diversity in quantitative surface analysis. Progress
in Physical Geography 24, 120.
Pirotti, F., Tarolli, P., 2010. Suitability of LiDAR point density and derived landform
curvature maps for channel network extraction. Hydrological Processes. www.
interscience.wiley.com doi:10.1002/hyp.7582. Published online in Wiley InterScience.
Pitty, A.F., 1971. Introduction to Geomorphology. Methuen, London. 526 pp.
Quine, W.V., 1981. Theories and Things. Belknap, Harvard U.P, Cambridge, Mass. 219 pp.
Rasemann, S., Schmidt, J., Schrott, L., Dikau, R., 2004. Geomorphometry in mountain
terrain. In: Bishop, M.P., Shroder, J.F. (Eds.), Geographic Information Science and
Mountain Geomorphology. Springer, Berlin, pp. 101145.
Reuter, H.I., Hengl, T., Gessler, P., Soille, P., 2009. Preparation of DEMs for
Geomorphometric Analysis, pp. 87120. Ch.4 in Hengl and Reuter, 2009 (q.v.).
Rose, J., Smith, M.J., 2008. Glacial geomorphological maps of the Glasgow region,
western central Scotland. Journal of Maps 2008, 399416http://www.journal
ofmaps.com 10.4113/jom.2008.1040.
Schmidt, J., Andrew, R., 2005. Multi-scale landform characterization. Area 37, 341350.
Schmidt, J., Dikau, R., 1999. Extracting geomorphometric attributes and objects from
digital elevation models semantics, methods, future needs. In: Dikau, R., Saurer,
H. (Eds.), GIS for Earth Surface Systems. Gebrder Borntraeger, Berlin, pp. 153173.

Schmidt, J., Hewitt, A., 2004. Fuzzy land element classication from DTMs based on
geometry and terrain position. Geoderma 121, 243256.
Schneider, M., Klein, R., 2010. Semi-automatic digital landform mapping. In Otto, J.-C.,
Dikau, R. (Eds.), Landform Structure, Evolution, Process, Control. Lecture Notes in
Earth Sciences 115, Springer, Berlin, pp. 3751.
Seijmonsbergen, A.C., Hengl, T., Anders, N.S., (in press). In Smith, M.J., Paron, P.,
Grifths, J. (Eds.). Geomorphological Mapping: a Handbook of Techniques and
Applications. Elsevier, Amsterdam.
Shary, P.A., 1995. Land surface in gravity points classication by a complete system of
curvatures. Mathematical Geology 27, 373390.
Shary, P.A., Sharaya, L.S., Mitusov, A.V., 2005. The problem of scale-specic and scalefree approaches in geomorphometry. Geograa Fisica e Dinamica Quaternaria 28,
81101.
Shary, P.A., Sharaya, L.S., Mitusov, A.V., 2002. Fundamental quantitative methods of
land surface analysis. Geoderma 107, 132.
Skidmore, A.K., 1990. Terrain position as mapped from a gridded digital elevation
model. International Journal of Geographical Information Systems 4, 3349.
Smith, M.J., Rose, J., Booth, S., 2006. Geomorphological mapping of glacial landforms
from remotely sensed data: an evaluation of the principal data sources and an
assessment of their quality. Geomorphology 76, 148165.
Smith, M.W., Bracken, L.J., Cox, N.J., 2010. Towards a dynamic representation of
hydrological connectivity at the hillslope scale in semi-arid areas. Water Resources
Research 46, W12540 doi:10.1029/2009WR008496.
Smith, M.J., Paron, P., Grifths, J. (Eds.), (in press). Geomorphological Mapping: a
Handbook of Techniques and Applications. Elsevier, Amsterdam.
Smith, M.J., Clark, C.D., 2005. Methods for the visualization of digital elevation models
for landform mapping. Earth Surface Processes and Landforms 30 (7), 885900.
Smith, B., Mark, D.M., 2003. Do mountains exist? Towards an ontology of landforms.
Environment and Planning B: Planning and Design 30 (3), 411427.
Smith, M.J., Pain, C.F., 2009. Applications of remote sensing in geomorphology. Progress
in Physical Geography 33 (4), 568582.
Smith, M.J., Wise, S.M., 2007. Problems of bias in mapping linear landforms from
satellite imagery. International Journal of Applied Earth Observation and
Geoinformation 9, 6578.
Smith, M.J., Rose, J., Gousie, M.B., 2009. The Cookie Cutter: a method for obtaining a
quantitative 3D description of glacial bedforms. Geomorphology 108, 209218.
Soille, P., Vogt, J., Colombo, R., 2003. Carving and adaptive drainage enforcement of grid
digital elevation models. Water Resources Research 39 (12), 13661378.
Speight, J.G., 1990. Landform. In: McDonald, R.C., Isbell, R.F., Speight, J.G., Walker, J.,
Hop, M.S. (Eds.), Australian Soil and Land Survey Field Handbook. Inkata Press,
Melbourne, pp. 957.
Strobl, J., 2008. Segmentation-based terrain classication. In: Zhou, Q., Lees, B., Tang, G.
(Eds.), Advances in Digital Terrain Analysis. (Lecture Notes in Geoinformation and
Cartography.). Springer, Berlin, pp. 125139.
Summereld, M.A., 1991. Global Geomorphology. Longman, Harlow. 537 pp.
Thomas, M.F., 1994. Geomorphology in the Tropics. John Wiley, Chichester. 460 pp.
Tobler, W.R., 1969. An analysis of a digitalized surface. In: Davis, C.M. (Ed.), A Study of the
Land Type, 86. University of Michigan, Department of Geography, MI, pp. 5976.
Report on Contract DA-31-124-ARO-D-456.
Tribe, A., 1991. Automated recognition of valley heads from digital elevation models.
Earth Surface Processes and Landforms 16, 3349.
van Asselen, S., Seijmonsbergen, A.C., 2006. Expert-driven semi-automated geomorphological mapping for a mountainous area using a laser DTM. Geomorphology 78,
309320.
Wilson, J.P., Gallant, J.C. (Eds.), 2000. Terrain Analysis: Principles and Applications. John
Wiley, Chichester, pp. 267294.
Wise, S.M., 2007. Effect of differing DEM creation methods on the results from a
hydrological model. Computers and Geosciences 33 (10), 13511365.
Young, A., 1972. Slopes. Oliver & Boyd, Edinburgh. 288 pp.