Вы находитесь на странице: 1из 37

CATCHMENT

DELINEATION AND CHARACTERISATION


A Review

by
Francesca Bertolo

C
C aatc
tchhm
meenntt
C
C hhaarraaccte
terrisa
isatio
tionn &
&
M
M ooddeellin
llingg

A n A c tiv ity o f th e E u ro L a n d sc a p e P ro je c t
S p a c e A p p lic a tio n s In stitu te (S A I)

Catchment Characterisation and Modelling


EuroLandscape Project
Space Applications Institute, Joint Research Centre
Ispra (Va), Italy

April 2000

Space Applications
Institute

Environment &
Geo-Information Unit

DG Joint Research Centre


European Commission

ii

Francesca Bertolo

For further information concerning the Catchment Characterisation and Modelling


activity or the EuroLandscape project you may contact:
Dr. Jrgen Vogt, JRC-SAI-EGEO, email: juergen.vogt@jrc.it or
Dr. Sten Folving, JRC-SAI-EGEO, email: sten.folving@jrc.it.
Or visit our website on URL: http://www.egeo.sai.jrc.it

JRC SAI EGEO

Catchment Delineation and Characterisation

iii

Table of Contents

Preface

1.

Introduction .................................................................................................. 1

2.

Digital Elevation Models .............................................................................. 2

3.

Catchment Delineation ................................................................................. 5


3.1

Defining Drainage From Raster Datasets ........................................ 6

3.2

The Problem of Flat Areas . ........................................................... 11

3.3

The Channel Source Definition ..................................................... 12

3.4

About Errors .................................................................................. 15

4.

River and Catchment Ordering ................................................................... 16

5.

Catchment Characterisation ........................................................................ 19

6.

Final Considerations .................................................................................... 22

7.

References .................................................................................................. 24

EuroLandscape CCM, April 2000

iv

Francesca Bertolo

JRC SAI EGEO

Catchment Delineation and Characterisation

Preface
Economic and environmental sustainability is one of the major goals of European policy.
One of the basic pre-requisites to meet these goals is a sound knowledge of the different
processes underlying economic and environmental evolution in the European territory. The
documentation of the current situation and the study of relevant processes are, therefore,
important issues for European institutions such as, for example, the Directorate General
Environment (DG ENV), the European Environmental Agency (EEA), the European
Statistical Office (EUROSTAT) or the Joint Research Centre (JRC) of the European
Commission.
In the frame of the 5th Framework Programme on Research and Technological
Development, the Environment and Geo-Information (EGEO) Unit of JRCs Space
Applications Institute (SAI) is aiming to construct such fundamental knowledge and
information through the implementation of the EuroLandscape project. EuroLandscape
(Geo-Information for Development and Environmental Monitoring) is aiming at assessing,
mapping and monitoring the European Environment, with special emphasis on the
sustainable management of natural resources, including forests, grasslands and water
resources.
The catchment, as a basic physical entity of the landscape, has gained increasing attention
in this context. Most processes related to the movement and quality of water are best studied
at the catchment or sub-catchment scale and many associated processes such as soil
erosion, mass movements, sediment transport, or land cover changes are strongly linked to
this spatial reference unit. The EEA, therefore, intends to set-up a comprehensive database
of catchment boundaries and river networks for the whole pan-European territory. Such
database should include information on the topology and ordering of both the catchments
and river stretches in order to allow for a detailed analysis of data on water quality and
quantity, which are collected through the EuroWaternet network of measurement stations.
This network is established by the EEA with the support of the national water authorities.
Eurostats sections for Geographic Information (GISCO) and for Water Statistics are
additional customers for the use and dissemination of such information. Together with the
planned physical and socio-economic characterisation of the mapped catchments, the
information will also be useful for the hydrological, geomorphological and socio-economic
modelling communities in order to put the results of a wide variety of models into a wider
geographical context.
Within EuroLandscape it is the Catchment Characterisation and Modelling (CCM) activity
to implement the catchment related work. CCM aims at a comprehensive mapping and
characterisation of catchments in Europe and at a subsequent modelling of key processes in
a set of representative catchments. This work is implemented at SAI, in close collaboration
with the EEA and with a network of experts in a variety of institutions throughout Europe.
This report has been written as a preparation to the European-wide mapping of catchments
and drainage networks. In parallel, different algorithms have been tested for selected
regions in Europe.

Ispra, April 2000

EuroLandscape CCM, April 2000

Jrgen Vogt
EuroLandscape CCM Leader

vi

Francesca Bertolo

JRC SAI EGEO

Catchment Delineation and Characterisation

1.

Introduction

The Catchment Characterisation and Modelling (CCM) activity of the


EuroLandscape project is aiming at a European-wide mapping of catchments and
drainage networks. The derived catchments shall then be characterised and classified
according to surface characteristics, land cover dynamics and run-off conditions. It
is in this context that this report on the state-of-the-art in catchment mapping and
characterisation has been prepared. In particular two aspects have been investigated:
how to automatically extract and map catchments and drainage networks from
digital elevation models (DEMs) and whether examples of catchment classifications
at regional level have been described in the literature.
The availability of a DEM with adequate spatial resolution and covering the whole
area of interest is a basic requirement to achieve the goals of CCM. It is evident that
the quality of the DEM is of high importance and that considerable attention has to
be paid to the errors that are introduced by the terrain model itself. This is related to
the problem of scale: does the geometric resolution of the DEM have any influence
on the precision of catchment boundaries and drainage channels and which degree of
error is introduced by using a given DEM? In connection to that, it is important to
highlight that the final database of European catchments will be the core of a
Geographic Information System (GIS). Since the propagation of errors plays a
significant role in GIS analysis it is important to have a clear understanding of how
to deal with them. Procedures have, therefore, to be implemented for evaluating the
reliability of both the input data and the final results (Chapter 2).
A further step relates to the available algorithms for delineating catchments and river
networks. It is not only important to know how to analyse the DEM, but also to
understand the limitations of each algorithm. The method that ideally fits the needs
of the EuroLandscape project must be computationally robust and efficient, and
must be able to cope with the most important problems that this kind of analysis
presents: DEM inherent altitude errors and the difficulty to recognise the pathways
of water in flat areas. Another interesting problem is the definition of the river
source area; in fact there is an intrinsic difficulty to define where exactly water starts
to form a river channel. Finally, there is a need for a theory to help in taking a
decision on the minimum source area that can be mapped from medium resolution
DEMs (Chapter 3).
Some attention needs to be given to the fact that, at the end, a large database has to
be managed. Within this database the geographic relationships between different
river networks or different catchments must be defined through an adequate
nomenclature or coding system. This nomenclature or coding system is necessary
for analysing the associated information (e.g. water quality, water quantity,
sensitivity to flooding) in a geographical context and for understanding the
environmental implications of political decisions. A specific chapter, therefore, is

EuroLandscape CCM, April 2000

Francesca Bertolo

devoted to investigate existing coding systems and to understand if it is possible to


implement a similar system at the European level (Chapter 4).
Finally, the question of catchment characterisation and classification involves a
problem of scale and of data availability. In this framework is important to
understand which are the most important parameters that characterise the landscape,
the hydrological response and the environmental behaviour of a catchment at
different scales, since CCM aims to obtain a database through which it will be
possible:
a) to derive a catchment typology and to select a set of representative catchments;
b) to aggregate catchments to identify homogeneous zones or groups of
catchment units at the pan-European scale;
c) to extrapolate model results from representative catchments to the entire panEuropean region.
Simultaneously, it is important to consider which data are available or can
reasonably be derived for the whole European territory (Chapter 5).

2.

Digital Elevation Models

Raster DEMs can be derived directly from stereo-photos or from satellite imagery
such as stereoscopic SPOT images, but are generally derived by interpolation of
scattered point elevation data, of contour lines, or of Triangulated Irregular
Networks (TIN).
The main limitation of a regular gridded DEM appears to be the fixed grid cell size.
Such a DEM cannot always accurately describe the topography, especially in
landscapes with varying complexity. Errors are introduced by the interpolation
procedure, whatever method is used and, in general, these errors are spatially autocorrelated. Moreover, in order to save memory, gridded elevation data are often
rounded to the nearest meter. In regions with gentle slopes this creates flat areas
with abrupt changes in altitude (similar to stairs).
All of these errors produce artefacts, such as pits and hummocks, that do not
correspond to real landscape features and they affect the derived quantities such as
slope-gradient and slope-aspect. While slope-gradient has about the same degree of
error as the original elevation data, slope-aspect errors are usually amplified during
calculation (Isaacson and Ripple, 1990; Bolstard and Stowe, 1994; Giles and
Franklin, 1996; Desmet, 1997).

JRC SAI EGEO

Catchment Delineation and Characterisation

These drawbacks have a clear impact for the extraction of the channel network and
for the catchment delineation, which are generally based on slope gradient and slope
aspect.
For most hydrological applications, the vertical resolution of a DEM is considered
satisfactory if the ratio of the average drop per pixel and the vertical resolution is
greater than unity. The average drop per pixel is defined as the elevation between a
pixel and the next in steepest descent (Thieken et al., 1999, Walker and Willgoose,
1999).
Scale and grid cell size influence the extraction of the channel network to a point
where the same method produces different results for the same area. In general, the
grid cell size dependency is introduced by the inability to accurately reproduce
drainage features that are at the same scale as the spatial resolution of the DEM. For
meandering channels, this results in shorter channel lengths and for networks with
high drainage density, it leads to channel and drainage area aggregation. In these
situations, the number of channels, the size of direct drainage areas and the network
pattern may depart considerably from the initial reference values (Wang and Yin,
1998). Garbrecht and Martz (1994) presented a sensitivity analysis on drainage
properties extracted from DEMs of increasing cell size and for several hypothetical
network configurations. On the basis of these results they found that a DEM should
have a grid cell area of less than 5% of the network reference area in order to
reproduce important drainage features with an accuracy of about 10%. The network
reference area is the mean area draining directly into the channel links of the
network.
The underlying data source used for deriving the DEM is a crucial factor. For this
reason, the aggregation of an accurate DEM is considered better than using a DEM
derived from maps at a lower scale (Thieken et al., 1999; Wolock and Price, 1994,
Walker and Wilgoose 1999).
Different studies highlighted that the importance of these issues is due to the fact
that the extent of the stream network and the length of the overland flow path
strongly influence hydrological modelling results (Wolock and Price, 1994; Zhang
and Montgomery, 1994; White and Running, 1994; Thieken et al., 1999).
Hussein and Schwartz (1997) presented a systematic strategy for improving the
quality of a DEM by including additional digital information on the geometry of the
stream network. The approach is based on a theory developed by Hutchinson (1988,
1989) for creating digital elevation models by combining point elevation data and/or
contour lines with a stream network. A unique feature of Hutchinsons approach is
an automatic drainage enforcement algorithm, which attempts to remove
spurious sinks in order to create a depressionless DEM. In addition, the algorithm
also modifies the elevation of grid points that conflict with a downstream decrease
in stream elevation. Control over the removal of spurious sinks and conflicting
elevation points is achieved using two tolerances: the first controls the maximum
EuroLandscape CCM, April 2000

Francesca Bertolo

difference between a sink considered for removal and the nearest pour point (i.e., the
local topographic minimum controlling a basin area). The second tolerance controls
the maximum allowable modification of grid points that conflict with the stream
network. The method of Hutchinson is incorporated in Arc/Info (ESRI) as a function
called Topogrid.
To correct the errors derived from elevation round-off, Nelson and Jones (1994)
proposed a smoothing filter. The weights of the filter matrix are determined using an
inverse distance squared function, so that the elevations of cells nearest to the central
cell will influence the result more than those of cells that are further away. A final
constraint is placed on the calculated elevation so that it is not adjusted by more than
one half the elevation resolution, to ensure that the terrain model is smoothed
without any loss of the original accuracy.
Only a few DEMs are available covering the whole pan-European area. One of them
is the so-called GTOPO30. It is a global DEM that was released by the USGS EROS
Data Center in 1996 (http://edcwww.cr.usgs.gov/landdaac/gtopo30/gtopo30.html).
The elevation values in GTOPO30 are regularly spaced at 30-arc seconds, which
corresponds to approximately one kilometre. They are derived from eight sources of
elevation information, including both vector and raster data sets. For most of Eurasia
the data source is the Digital Terrain Elevation Data (DTED) produced by the
National Imagery and Mapping Agency (NIMA, formerly the Defense Mapping
Agency), which is a raster topographic database with a horizontal grid spacing of 3arc seconds (approximately 90 meters). The generalization of the high-resolution
data to the 30-arc seconds horizontal grid spacing was conducted by calculating the
median value of the 100 full resolution cells corresponding to each cell of the new
DEM. For this reason the accuracy of the values obtained is the same as in the
original data set. The full resolution 3-arc seconds DTED has a vertical accuracy of
30 meters linear error at the 90% confidence level, which correspond to a RMSE
of 18 meters. To ensure that the DEM is able to reproduce the correct movement of
water across its surface, the DEM is processed to remove elevation anomalies that
can interfere with hydrologically correct flow (Verdin and Jenson, 1996).
Another high resolution DEM of Europe is distributed by GAF mbH, Germany,
according to a sales agreement with GEOSYS/MONA PRO Visual Media, France.
The MONA PRO DEM covers 22 countries in Europe and is available with the grid
cell sizes of 75 m, 100 m and 250 m. The altitude precision (according to GEOSYS)
is about 3,5 5 m in relatively flat terrain, and 12 15 m in very steep mountains
(GAF Products and Services, http://www.gaf.de/gaf04.htm).
The X-SAR/SRTM Shuttle Radar Topography Mission flown on the Space Shuttle
in February 2000 will result in another potentially interesting DEM. This joint
mission of the US National Imagery and Mapping Agency (NIMA), the US National
Aeronautics and Space Administration (NASA), the German Aerospace Centre
(DLR) and the Italian Space Agency (ASI) had the objective to use C-band and Xband interferometric synthetic aperture radars (IFSARs) to acquire topographic data
JRC SAI EGEO

Catchment Delineation and Characterisation

over 80% of Earths land mass (between 60 degrees North and 56 degrees South)
during the 11-day Shuttle mission. Within 18 months after the mission the data will
be processed into digital topographic data with a 30 x 30 m spatial sampling rate.
The expected accuracy of the resulting data is given as 16 m absolute vertical height
accuracy, 10 m relative vertical height accuracy and 20 m absolute horizontal
circular accuracy at a 90% confidence level. Worldwide data will be distributed at a
spatial resolution of around 100 m. When available, the X-SAR/SRTM DEM will be
the first continuous high-resolution product, which has not been mosaiced from data
derived from differing sensors, formats and dates. (http://www.jpl.nasa.gov/srtm/;
http://www.dlr.de/srtm/).
Higher resolution DEMs could also be collected from national databases. Although
these DEMs are generally of high quality they are usually very expensive. In
addition, it is generally difficult to join adjacent DEMs that come from different
sources and have been produced in different ways. Finally, with increasing spatial
resolution there is an increasing need of storing capacity and of computational
power. Clearly, these problems limit the use of such data within the CCM project. A
homogeneous DEM, covering the whole area of interest should be the preferred
option.

3.

Catchment Delineation

The last years saw a general recognition of the catchment or the drainage basin as
the most significant surface unit in environmental studies. Traditionally catchment
boundaries have been manually derived from topographic maps, a labour-intensive
activity. This limitation has changed after the introduction of Digital Elevation
Models (DEMs).
Even though methods for delineating catchment boundaries and flow paths from
contour lines (Moore and Grayson, 1991) and triangulated irregular networks (Jones
et al., 1990; Palacios-Velez and Cuevas-Renaud, 1986) provide reliable results, they
require extensive data storage and computation time. Grid cell elevation models
have advantages for their computational efficiency and the availability of
topographic databases (Sabbagh et al., 1994). Therefore, they have seen widespread
application for analysing hydrological problems.
There is only one method, which does not require a DEM; it is based on an
automated river network overlay (Sekulin et al., 1992). This approach needs a welldefined river network database, in which the basic unit is a river stretch (link)
defined as the river length between two nodes. Each cell of a grid is allocated to
river stretches using a shortest distance algorithm. Boundaries of hydrometric
areas, coastlines, and boundaries of the catchment area above the gauging stations
can be used with a point-in-a-polygon algorithm to give added precision in the
allocation phase. Grid cells are then accumulated upstream of river stretches using a
EuroLandscape CCM, April 2000

Francesca Bertolo

down-network travel technique. This procedure has been proposed in the ERICA
(European Rivers and Catchments) project of the EEA (Flavin, 1998).
The catchment delineation is a two-step procedure: drainage patterns have to be
recognized before the boundaries between different catchments can be inferred.
3.1

Defining Drainage Networks From Raster Datasets

Three main approaches to the automated recognition of valleys and drainage lines
from raster DEM can be identified (Tribe, 1992):
A. The recognition of individual DEM cells as valley cells, where a cell is
classified as a valley if some of the cells neighbours are higher;
B. The assignment of drainage directions to each DEM cell and the use of this
information for the derivation of a drainage network;
C. Two-step methods based on a combination of approaches A and B.
They will be described in the following sections. Finally, a few remarks concerning
the scale properties and fractal geometry of drainage networks will be given in
section D.
A. The recognition of individual DEM cells as valley cells
In the first approach individual DEM cells are identified as valley cells by
comparing the heights of each neighbour of the cell in turn with that of the cell. The
methods of comparison are based on different concepts:
graphs constructed from the elevations of cells neighbours (If the cell represents
a valley cell this graph will conform to a particular configuration);
the connectivity number and the coefficient of curvature are calculated from the
cells neighbours;
comparison of the elevations of the neighbour cells in predefined directions
(Peucker and Douglas, 1975).
The methods described produce noise and valleys, which extend too far up-valley
because they are based on the concept of higher than. Higher could be only 1 m,
meaning that any cell representing a hardly discernible or a local depression can be
classified as part of the valley network. Generally the network is discontinuous and
some procedure is needed to connect the different valley segments and finally to
derive the channel network by thinning.

JRC SAI EGEO

Catchment Delineation and Characterisation

B. The assignment of drainage directions to each DEM cell.


Mark (1984) and OCallaghan and Mark (1984) noted the discontinuities produced
by the previous methods and proposed an algorithm to produce a continuous
network. The algorithm, known as the D8 algorithm, is based on the flow of water
over terrain along lines of steepest slope and is a computerized version of a manual
method of catchment area measurement (Speight, 1968). Each cell is considered to
drain to whichever of its eight neighbours has the steepest downslope from it. This is
not always the lowest neighbour, since the height differences for the four diagonal
half-neighbours of a point must be divided by the grid spacing multiplied by the
square root of two. Initially, each cell may be considered to produce a unit quantity
of runoff; this runoff is then carried downslope in accordance with drainage
directions of the grid cells. Then, whenever the runoff in a cell exceeds some
threshold, the cell is considered to be part of the drainage network. The principal
limitation in the method is that each cell of the DEM has to have a drainage
direction assigned to it.
The D8 algorithm appears in many works, and in particular in Jenson and Domingue
(1988) on which the ARC/Info tools for catchment delineation are based. The
biggest drawback of this method is the fact that it represents only convergent flow.
To overcome this limitation in flow direction assignment different approaches have
been proposed.
Fairfield and Leymarie (1991) proposed to introduce a stochastic rule in order to
follow more closely the aspect of the slope to avoid the fact that in the D8 algorithm
the flow is discretized to only one of eight directions, separated by 45. The
disadvantage of this new procedure is that the result is not exactly reproducible
because of its randomness.
Some authors have proposed that flow must be partitioned between different pixels.
Freeman (1991) allocates flow to each lower neighbour in proportion to an exponent
p of the slope. According to his results, a value of p = 1.1 is appropriate. These
methods have the disadvantage that flow from a pixel is dispersed to all
neighbouring pixels with lower elevation. Therefore the contributing area of a pixel
does not include any full pixel but instead is composed of portions of different pixels
and is discontinuous.
Costa-Cabral and Burges (1994) oppose to most current models because of their
point source representation of flow generation and the resulting one-dimensional
representation of flow paths; as an alternative they presented an elaborate set of
procedures which model downslope flow in two dimensions in well-defined flow
tubes. Flow at one point is in the direction of maximal surface slope. For planar
pixels, if the flow direction is parallel to the grid orientation, the exit portion of the
boundary is a single full boundary segment. If the flow direction is not parallel to the
grid orientation, the exit portion of the boundary consists of two full adjacent
boundary segments.
EuroLandscape CCM, April 2000

Francesca Bertolo

Tarboton (1997) tries to reduce dispersion by dividing the flow between one or two
downslope pixels. The flow direction is a continuous quantity between 0 and 2 that
is determined in the direction of the steepest downwards slope on the eight
triangular facets formed in a 3 x 3 pixel window centred on the pixel of interest.
Where the direction does not follow one of the cardinal or diagonal directions, the
upslope area is calculated by distributing the flow from a pixel between the two
downslope pixels according to how close the flow angle is to the direction of the
pixel centre. Unresolved flow directions, in flat areas or depressions, are resolved
iteratively by making them flow toward a neighbour of equal elevation that has a
flow direction resolved. A suite of programs for the Analysis of Digital Elevation
Data
(TARDEM)
from
David
Tarboton
is
freely
distributed
(http://www.engineering.usu.edu/dtarb/).
For all of those methods depressions (pits) cause serious problems, but also in flat
areas the assignment of a flow direction is not obvious. While closed depressions
and flat areas in a DEM may represent real landscape features, they are more often
artefacts that arise from errors in the input data, interpolation procedures, and the
limited horizontal and vertical resolution of the DEM (Mark, 1984; Jenson and
Domingue, 1988; Tribe, 1992; Martz and Garbrecht, 1998; Zhang and Montgomery,
1994).
The correction of spurious pits has been conducted principally with two strategies.
The first strategy attempts to remove depressions by smoothing the DEM data
(Mark, 1984). Objections to this method are threefold:
It does not distinguish between natural and spurious pits.
It fails to remove all spurious pits, in particular deeper ones.
A loss of significant information is evident after smoothing.
The second strategy is to fill depressions by increasing the values of cells in each
depression to the value of the cell on the depression boundary with the lowest value
(Jenson and Domingue, 1988). First, the algorithm finds the pits outflow point: that
cell on the pits boundary where water would flow out of the pit if it were filled with
water. Then the heights of all cells in the pit, lower than the outflow, are changed to
the height of the outflow. This creates a flat area over which drainage directions can
be assigned.
Fairfield and Leymarie (1991), among others, suggested to treat the pits as if they
were real depressions, and to find the lowest point, the pass, from which water could
flow out of the pit basin. This is achieved changing the directions of flow on the
path between the pass and the low point of the pit. Water is figuratively made to
climb up the side of the basin, which is unrealistic, but the DEM is not changed.

JRC SAI EGEO

Catchment Delineation and Characterisation

Martz and Garbrecht (1998), from the consideration that a pit can be produced also
by elevation overestimation errors, proposed a breaching algorithm which
simulates breaching of the outlet of closed depressions to eliminate or reduce those
expected to have been produced by elevation overestimate. It evaluates the local
outlet of each closed depression in a DEM to determine whether the elevation of one
or two cells at the outlet could be lowered to eliminate or reduce the size of the
depression without reversing the direction of overland flow across the outlet.

C. Two-step methods
Several researchers have developed two-step procedures for extracting channel
networks. Start cells or disconnected valley segments are first identified using a
method similar to, or based on the concept of higher than, and these are then
grown, usually down the line of steepest slope, to give continuous drainage lines
(Band, 1986; Riazanoff et al., 1992; Lammers and Band, 1990; Skidmore, 1990;
Yoeli, 1984; Smith et al., 1990).
Yoeli (1984) proposed a valley line finding algorithm that, first of all, finds
elevation minima on the DEM using a spline curve. A continuous search for the next
lower neighbour is then conducted in a sampling square of twice the grid interval
around the current last point of the valley line. Valley lines start from the highest
minimum which, at this stage does not lie in an existing valley line, and finish when
one of three possible situations is reached: joining another valley line; flowing into a
lake or sea; reaching the edge of the DEM.
Band (1986) in a first step marks convex- and concave-upward points as ridge and
stream points, respectively, then the procedure searches for segment ends; new
pixels are added downstream to the segments using a maximum descent algorithm
until another stream segment is encountered. The final result is obtained by thinning
the image to single pixel-width lines. Bands stream network and sub-catchment
extraction algorithms, along with the production of topological codes describing
their structures have been released as part of GRASS, a US Army Corps of
Engineers public domain GIS (http://www.geog.uni-hannover.de/grass/).
Smith et al. (1990) describe a two-step method based on a procedure first proposed
by Haralick (1983). A cubic surface is fitted to the neighbourhood of each pixel. A
pixel is defined as a valley pixel if the first directional derivative has zero crossings
in a direction in which the second-directional derivative has positive extreme values.
In a second step the procedure applies knowledge about drainage networks to
integrate these probable valley pixels into a network of single-pixel-width lines
satisfying the constraints imposed by a binary tree model.
Riazanoff et al. (1992) and Chorowicz et al. (1992) first identify saddles, points that
divide two groups of pixels in a neighbourhood which have higher elevation than the
saddle. In their method these researchers use two layers: one is the DEM and the
EuroLandscape CCM, April 2000

10

Francesca Bertolo

second is a virtual image of the network represented by segments. From each saddle,
a segment is initialised in all the possible directions (generally not more than two);
segments are grown in the virtual image in the direction of maximum slope, until
another segment, or another saddle, or the border of the DEM is reached. The
resulting network is very dense.
Meisels et al. (1995) propose a two-stage algorithm of multilevel skeletonisation of
a DEM, followed by a process of enumeration, mainly to eliminate loops in the
extracted drainage network. Depressions are filled to the surrounding elevation level
before the extraction of the drainage network. The main algorithm extracts pixels
that lie on high curvature contours starting from pixels of maximal elevation,
elevation-level by elevation-level; the selection is based on a condition for a large
enough number of higher elevation pixels in the immediate neighbourhood of a pixel
belonging to the elevation currently being processed. The second stage uses a
complementary local condition of connectivity and connects all the pixels of the
flow path.
In general, these methods suffer from the same problems as the methods described
in the first section; they are effective only when the drainage network is well defined
by the local surface properties, which can be derived from the DEM.

D. Scale properties and fractal geometry of channel networks


In the context of the studies on scale properties of channel networks, several
researchers have been interested in the fractal geometry of individual streams and
channel networks as a whole (Hjelmfelt, 1988; Tarboton et al., 1988; LaBarbera and
Rosso, 1989). The fractal nature of river networks manifests itself in two ways: on
the one hand, the plane pattern of an individual watercourse has fractal geometry
and, on the other hand, fractal properties are also characteristics of the branched
pattern of river networks (Da Ros and Borga, 1997). As a conclusion from all these
works, Roth et al. (1996) propose a global approach to the problem of drainage
network identification. All empirical evidence and the theoretical description of the
hydrodynamical and morphological conditions, which are expected to hold in the
streams constituting the effective drainage network, can be expressed in the form:
AS = constant

(1)

Where A is the drainage area, S is the local slope and is a constant ranging from
0.44 to 0.5. The practical application of this relation for the extraction of the channel
network is often limited by the low accuracy of the elevation data; in particular the
evaluation of the local slope is a critical issue. The resulting network is often formed
of non-connected streams and therefore this approach is only used as a theoretical
basis for the design of filtering procedures.

JRC SAI EGEO

Catchment Delineation and Characterisation

11

Another approach, which LaBarbera and Roth propose, starts from the analysis of
the altimetric properties of the area that contribute to a given site. Under the
hypothesis of self-similarity of stream slopes, as described by the Hortonian slope
ratio, the relation Sl Sa can be assumed between the local slope, Sl, and the average
slope in the subcatchment draining in the point being studied, Sa. Introducing:
Ha = (1/n)i (Hi-Hs)

(2)

as the relative elevation of the subcatchment draining in the selected point, we


obtain that:
Sa Ha/L
(3)
where n is the number of pixels in the subcatchment, Hi is the elevation of a pixel in
the subcatchment, Hs is the elevation in the site studied and L is a linear measure of
the subcatchment size. Introducing the fractal dimension of single rivers, d, a
relation with the contributing area has been defined (Rosso et al., 1991):
L Ad/2

(4)

Ha Ad/2 Sl

(5)

Equation 5 provides a link between the relative elevation of the subcatchment at a


given site, Ha, the contributing area, A, the fractal dimension of single rivers, d, and
the local slope, Sl. Moreover, the structure on the right-hand side of this equation is
the same as the general equation 1. The application of this approach leads to a wellconnected and coherent network; the spatial variation of the drainage density is well
reproduced with a high drainage density in the mountain regions that tends to
decrease towards the alluvial areas, characterized by low slope values, and a very
low density in the plains.

3.2

The Problem of Flat Areas.

In a DEM, there is a well-known difficulty to discriminate between flat areas


drained by incised channels and truly flat areas that carry water as sheet flow. It is a
general problem that follows from the horizontal resolution of the DEM: it is clear
that when the size of a drainage feature is much smaller than the grid cell size of the
DEM, the channels cannot be captured by the DEM. However assigning drainage
directions to cells where there are two or more possible choices, and to flat areas, is
a general problem for all the methods described previously.
Generally, following Jenson and Domingue (1988), flat area cells adjacent to other
cells with a defined flow direction are identified. These flat area cells are then
assigned a flow direction, pointing to the nearest adjacent cell with a defined flow
direction. This is repeated until all flat area cells are assigned a flow direction. This
EuroLandscape CCM, April 2000

12

Francesca Bertolo

kind of approach constrains the flow path to remain within the flat area and allows
the possibility of multiple outlets. However it often produces unrealistic, parallel
flow patterns.
Other researchers proposed to infer flow paths in flat areas from the surrounding
topography. Tribe (1992) suggests defining a main flow path through the flat area
and directing other flow paths towards this main path. Unfortunately, because the
main flow path is defined along the shortest path between the inflow and the nearest
outlet, it is possible for the main flow path to pass through areas of higher elevation
to reach the outlet.
Garbrecht and Martz (1997) and Martz and Garbrecht (1998) describe an algorithm,
which is a core component of the freely available TOPAZ (TOpographic
PArameteriZation) landscape analysis tool (http://duke.usask.ca/~martzl/topaz/).
The method is based on the recognition that in homogeneous natural landscapes the
drainage is generally towards lower terrain while simultaneously being away from
higher terrain. Such a drainage is achieved by imposing two gradients on the flat
surface: one towards lower terrain which draws flow to the nearest downslope
outlet, and a second which forces flow away from higher terrain.
Mackay and Band (1998) recently proposed to first identify flat features (e.g. lakes
and relatively flat areas) on the DEM for which slope tracking is likely to fail.
Contiguous groups of flat areas are formed into labelled regions by using local
region growing. The labelled regions are then classified into water bodies or land
areas using supervised classification of remotely sensed imagery. Different cellbased algorithms can be associated with each class of flat feature: for lakes the goal
is to deliver the total upslope area contributed to the lake and the lake itself to the
lake outlet, for other flat areas it is necessary to optimise the slope threshold needed
to define the flow path (Band, 1986). Since actual flow of water through a lake
requires additional information, a lake boundary-following procedure is used: each
grid cell along the boundary acts as a depression point for catchments that drain into
it. The contributing area assigned to the lake outlet cell is the total area along the
boundary plus the area of the unmarked cells in the interior of the lake; this retains
the topology of the land-lake features, but eliminates cell-to-cell flow within the lake
itself.

3.3

The Channel Source Definition.

The delineation of catchments is an important aspect in hydrological modelling and


is closely related to the definition of channel networks. Physically based
hydrological models must distinguish between overland runoff and water that flows
in channels. On the other hand a catchment outlet must be on a river course. It
appears clearly that the correct delineation of a catchment is closely related to the
correct extraction of its channel network.
JRC SAI EGEO

Catchment Delineation and Characterisation

13

Hydro-physically, the channel network represents those points at which runoff is


sufficiently concentrated for fluvial processes to dominate over slope processes
(Mark, 1984). This mechanism is basically driven by the topography; therefore a
different kind of information is needed to define the conditions of this transition,
which introduces an indetermination in the methods available.
The correct automatic derivation of the channel network is an unsolved problem.
Field investigations showed that the source area above the channel head decreases
with an increasing local valley gradient of the slopes, except in locations where
bedrock properties control the channel head locations (Montgomery and FoufoulaGeorgiou, 1993). Although steeper channel heads are, at least, partially controlled
by slope instability, landslide instability alone is not sufficient for canalisation.
Critical sheer stress is a dominant control factor on the extent of the channel network
when saturation overland flow is significant, but it is difficult to quantify this in the
field.
The most common method of extracting channel networks from DEMs is to specify
a critical support area that defines the minimum drainage area required to initiate a
channel (Mark, 1984; Band, 1986; Jenson and Domingue, 1988). In practice, this
threshold value is often selected on the basis of visual similarity between the
extracted network and the blue lines depicted on topographic maps. Tarboton et al.
(1989, 1992) propose a method to find the critical value of the contributing area
from the scaling diagram of slope versus contributing area for individual grid cells
within a catchment. An example is given in figure 1.
Contributing Area (m2)
1.E+04
1.E+00

1.E+05

1.E+06

1.E+07

1.E+08

1.E+09

1.E+10

Fluvial process
(Channelized pixels are
considered those with A > At)

Local slope (m/m)

1.E-01

1.E-02

Flu
v

Hillslope
processes

ial S A
sca
ling
line

- 0.5
1.E-03

At = 5 Km2 = 80 pixels

Dd A0.5
Figure 1: Diagram of local slope versus contributing area, derived for the Po
catchment in northern Italy from a DEM with 250 m grid-cell size.

EuroLandscape CCM, April 2000

14

Francesca Bertolo

The theoretical argument for using an area-threshold is based on the hypothesis that
the channel network extends up to the point where unstable fluvial sediment
transport processes change to stable diffusive hill-slope processes. Several
researchers showed that the change of dominance from hill-slope to fluvial
processes is equivalent to a change in the slope versus contributing area diagram
from a positive-gradient (dS/dA > 0) at small values of the contributing area, to a
negative-gradient (dS/dA < 0) at larger values of the contributing area. Tarboton et
al. (1989, 1992) propose to use the size of the contributing area that corresponds to
the change in the scaling response, as the critical support area for finding the
extension of the channel network. The break in slope-area scaling, however, seldom
goes from a positive to a negative gradient as predicted by the theory, but appears
instead as an inflection from a low to a high negative gradient in the slope-area
diagram (see figure 1).
Montgomery and Dietrich (1988, 1992) demonstrate that channel heads lie at a
transition between channelled and unchannelled portions of the landscape but that
for any given slope the size of the source area may vary by as much as an order of
magnitude. In other words the threshold area is not constant in a basin but it is a
function of the local valley slope (the slope immediately upstream of the channel
source in the unchannelled valley). Their research has identified an empirical
relationship (power law) between threshold area and slope:
Ath = CS-

(6)

where C and are constants empirically determined from field data and S is the
local valley slope. Identification of an appropriate value for C is a major impediment
for implementing this model, as this parameter should vary with both rainfall and
critical shear stress of the ground surface, the latter reflecting both soil properties
and the type and density of vegetation cover (Montgomery and Foufoula-Georgiou,
1993).
Using a slope-dependent threshold, drainage density is greater in steeper portions of
a catchment, which generally corresponds to the situation found in natural
landscapes. For the case of a catchment with little spatial variability in slope, the
constant threshold and the slope-dependent threshold methods converge and predict
similar channel networks for the same mean source area. (Montgomery and
Foufoula-Georgiou, 1993). However the area-threshold and the slope-area-threshold
criteria need not to be conflicting and they may be even combined in a single
framework, or be considered as a representation of processes over different
geomorphic time scales (Ijjasz-Vasquez and Bras, 1995).
The general conclusion of the majority of works on the source area location is that
even limited field data collection on the drainage area-slope relation for channel
heads, is the best method for determining appropriate values of parameters defining
channel network extent (Montgomery and Foufoula-Georgiou, 1993; Helmlinger et
al., 1993).
JRC SAI EGEO

Catchment Delineation and Characterisation

15

All the analyses described above refer to DEMs of maximum 30 m grid cell size; it
is stated several times in those works that the correct solution to the question of the
channel initiation needs high resolution DEMs, able to depict the landscape in a
consistent way.
At lower resolution it is more appropriate to analyse the problem from another point
of view. The fundamental assumption is that runoff is generated uniformly over the
basin and that the geomorphologic or erosional response to that runoff is spatially
homogeneous regarding channel initiation. Under these conditions the threshold
contributing area can be used to represent the complex interaction of factors such as
geology, vegetation, soils and topography, which control the initiation and
maintenance of a channel network. At a regional level, where the conditions of
uniformity may not be assumed, it is better, either to allow the threshold
contributing area to vary between different geomorphic regions, or to include
physical parameters that underlie the spatial variability of the geomorphology
(Martz and Garbrecht, 1995).
The importance of the selection of the threshold area is related to the high impact
that this value has on the morphometric properties (such as drainage density, length
of drainage paths, statistics of external and internal links) and scaling properties
(such as Hortons ratios and fractal dimensions) of a channel network (Helmlinger et
al., 1993; Moussa and Bocquillon, 1996; Da Ros and Borga, 1997).

3.4

About Errors.

As stated frequently before, the final result of the automatic delineation of the
catchments is influenced by many sources of errors. Selecting a DEM of good
quality and paying attention to the algorithm used can minimize some of the
possible errors a priori. Other errors need the definition of some criteria that help to
limit their influences. Finally, some errors are inherent to all the procedures and
need to be clearly evaluated and quantified.
Generally, the evaluation of the results of catchment delineation and network
extraction procedures is made by visual comparison with existing vectors. It is,
however, difficult to quantitatively evaluate the results. Generally, quantitative
comparisons are made on the total area included in the catchment boundaries.
Miller and Morrice (1996) proposed to evaluate the reliability of the catchment
boundaries using the rate of change of height and aspect in the neighbourhood of the
borderlines. In order to assess the sensitivity of the boundary location to errors in the
DEM, they proposed to add an error with normal distribution and fixed standard
deviation to the elevation values and to quantify the areas that would be re-allocated
between catchments. It is supposed that the sensitivity to DEM errors is best
EuroLandscape CCM, April 2000

16

Francesca Bertolo

represented by the area of land that is subject to re-allocation as a proportion of the


catchment area.

4.

River and Catchment Ordering.

In order to construct a geographical database of all the catchments of Europe, a


crucial point is to define a system for the codification of the catchments and related
channel networks.
This is a necessary step in order to maintain the topology of the hydrological system
and is useful to facilitate the access to the database. A unique code that identifies
each catchment and its river network is necessary for the structure of a database. But
it is also important for describing the relationships that exist within a hydrological
region, or between mainstreams and tributaries, or between headwaters, drainage
system and outlet, etc.
Each country has its own system of coding the national river network. Generally the
network itself influences the coding system through characteristics such as the
drainage density, the shape of the drainage system (i.e. coastal drainage or no sea
outlet), or the mean drained area. It is difficult to find a method that fits to all the
possible combinations that arise on a continent.
Another relevant aspect is the number of digits (alphanumeric elements) that is
necessary to completely describe a catchment: at a national level this is less
important than at a continental scale where thousands of catchments have to be
coded.
In the United States the Water Resources Division of the US Geological Survey has
its Hydrologic Units System (Seaber et al., 1987), which divides the Nation in 21
major geographic areas, or regions, composed of 222 sub-regions. A sub-region
includes the area drained by a river system, a reach of a river and its tributaries in
that reach, a closed basin, or a group of streams forming a coastal drainage area. The
third level of classification subdivides many of the sub-regions into accounting units
and, furthermore, the fourth level in cataloguing units, which are the smallest
elements in the hierarchy of hydrologic units. An eight-digit code uniquely identifies
each of the four levels of classification within four two-digit fields. The first two
digits identify the region, the next two digits the sub-region, the next two digits the
accounting unit and the last two digits the cataloguing unit.
In France the Environment Ministry has defined a coding system in 1991, which is
based on a code of eight digits. The first four digits indicate the hydrographic zone,
which refers to a hydrological classification of the country in four levels: regions (1st
level), sectors, sub-sectors and zones. There are six regions that represent the most
important river basins of the country. Each region is subdivided in not more than ten
JRC SAI EGEO

Catchment Delineation and Characterisation

17

sectors, each sector in not more than ten sub-sectors that have not more than ten
zones inside. The next three digits (5, 6, 7) refer to a single entity which can be
one out of five different inland waters: river, tributary, artificial channel, lake,
wetland, or coastline. The last digit refers to the class of inland water. An additional
parameter, the kilometric point (pk), defines the position of a specific point along or
in the border of an entity. It is used to concatenate each entity with the neighbouring
entities and to describe changes in the entity, like different regulations for water use
or for fishing, different fish species, or different owners, etc.
A different approach is used in the Baltic Sea Region GIS, Maps and Statistical
Database developed by United Nations Environmental Program - Global Resource
Information Database (UNEP/GRID) Arendal (Norway) in collaboration with
Institutes in Sweden (http://www.grida.no/prog/norbal/baltic/ welcome.htm). In this
database there are only 81 sub-basins, in total forming the seven major catchments
that define the Baltic Sea drainage area. The sub-basins all have an outlet to the sea
or are coastal drainage areas; portions of the Baltic Sea are considered sub-basins.
Therefore the sub-basins are numbered sequentially in clock-wise order beginning
from the northern catchment (Bothnian Bay); jumping in the numbering scheme can
occur when a portion of sea is encountered: a number ending by nine is always
assigned to it. An additional parameter is linked to each sub-basin indicating the
major catchment it belongs to.
The Catchment Database for Sub-equatorial Africa (Verheust and Johnson, 1998) is
another example of a regional database. The catchment delineation for subequatorial Africa was done in 1997 based on a digital elevation model and the river
layer of the Digital Chart of the World. The database was developed for users who
do not have access to high-end GIS packages and can be used as a standalone
cellular database or in combination with simple mapping packages. The database
holds 1157 catchment polygons; the ordering was done manually and names
assigned using a map of Central and Southern Africa from J. Bartholomew & Sons
LTD at a 1:5,000,000 scale (1988). At the same time, a parameter was added to each
polygon, which identifies the next catchment downstream. With a repetitive linking
of this parameter to the catchment identifier a downstream sequence is generated
and stored in a related database. Based on the created downstream sequence an
upstream sequence was also generated and stored in another related database. The
first record of this sequence refers to the outlet (ocean, internal basin, or edge basin)
and the second refers to the lowest catchment, which drains to the outlet. This lowest
catchment is the outlet area of the principal river of the network, and gives the name
to a megabasin. Two other parameters associated to each catchment complete the
description of the topology of the drainage system in this database: the level,
indicating the number of catchments that separate it from the outlet, and the
megabasin name.
Verdin and Verdin (1999) propose a reference system that at once uniquely
identifies and indicates the spatial nature of a hydrographic basin, with the aim to
have a simple and globally applicable method of coding. The system proposed is
EuroLandscape CCM, April 2000

18

Francesca Bertolo

founded upon concepts first articulated by Otto Pfafstetter, an engineer within the
Departamento Nacional de Obras de Saneamento, a civil works agency of the
federal government of Brazil (Pfafstetter, 1989 as cited in Verdin and Verdin, 1999).
This system is designed to exploit features of the base-10 numbering system: the
ordinal nature of digit values indicates the relative upstream/downstream position,
while the binary trait of being alternately odd and even indicates the possibility of a
stream to be on or off the main channel. In order to apply the method to a river, it is
necessary to distinguish between the main stem and tributaries: the main stem is
defined here as the watercourse that drains the greater area. The area drained by a
tributary is called basin while the area directly drained by the reach of the main stem
lying between two tributaries is called inter-basin.
Subdivision of the area drained by a major river into coded basins and inter-basins is
based on the convention to increase ordinal values from downstream to upstream,
and to assign odd digits to inter-basins and even digits to basins. A zero digit is
reserved for areas of closed drainage. There will be a large number of candidate
tributary basins to be delineated, but only four digital values are available per level:
the four tributaries with the greatest drainage areas are assigned a value. However,
the value assigned to a basin or inter-basin depends on its topological position and
not its area. Values are assigned in the order in which they are encountered along the
river course from the outlet to the source. Any basin or inter-basin can be further
subdivided by simply applying the same rules to its internal area adding a digit for
each level of sub-division. The appeal of the approach stems from its economy of
digits, the topological information that the digits carry, and its global applicability.
Simple rules to check digits with tests of odd or even, and less than or
greater than, can quickly isolate areas of interest for a particular investigation.
A coding system for the territory of the European Union has been proposed in the
ERICA (European Rivers and Catchments) project of the EEA (Flavin et al., 1998).
The aim of this coding system is to provide explicit information as to areas draining
to a given sea/ocean or coastal stretch, identification of all areas above or below a
given point, and the size of a catchment. This result is achieved through a code that
has:
Two digits to identify the sea that the catchment drains to;
Three digits to identify the coastal position of the area drained;
A series of two digits for nested catchments;
A single character indicating the catchment size.
For the first two digits of the code the International Hydrographic Bureau system
(1953) is used. This is a widely used system of coding for seas and oceans.
The second group of digits, the marine border code, can assume two kinds of values:
an even number indicates a true river outlet to the sea or the border between two
oceans (water-water boundary); an odd number indicates a stretch of coast between
two river outlets or two oceans (water-land boundary). The 499 (total number of
even numbers between 1 and 999) most significant rivers are numbered sequentially
JRC SAI EGEO

Catchment Delineation and Characterisation

19

from north to south and from west to east. To identify the most significant rivers it is
proposed to use the catchment area or, if this is not available, the total stream length
draining to each mouth.
A drainage area can be subdivided in smaller drainage areas according to the
bifurcating structure of rivers. At each level of subdivision the 49 (the total number
of even numbers between 1 and 99) most significant tributaries are numbered
sequentially from the outlet to the source. The inter-catchment areas between
significant tributaries will take the odd numbers. To identify the most significant
tributaries the drained area can be used or, if this is not available, the total stream
length. For each level of subdivision two digits are added. This subdivision
procedure can apply also in inter-catchment areas and coastal drainage areas.
The last element in the code is a character, the area band code, which indicates the
size of the catchment essential for comparing like-with-like at a European scale.
This code refers to ranges of size to be defined.

5.

Catchment Characterisation.

The Catchment Characterisation part of CCM aims at a European-wide


characterisation and classification of catchments according to surface characteristics,
land cover dynamics and run-off conditions. Subsequently, run-off, soil erosion and
sediment dynamics will be modelled for selected (representative) catchments in
order to assess the risk of land degradation and the needs for environmental
protection. Within this context, emphasis is placed on characterising the landscape
and modelling landscape processes. As such the approach extends beyond more
traditional hydrological characterisation and modelling and should provide the basis
for:
1. a comprehensive catchment classification and thus the production of typologies
and the selection of a set of representative catchments;
2. the aggregation of catchments to identify homogeneous zones or groups of
catchment units at the pan-European scale;
3. the extrapolation of model results from representative catchments to the entire
pan-European region.
Catchment characterisation, therefore, should be based on a set of fundamental
multipurpose parameters capable of representing both the diversity and the key
features of the pan-European landscape.
Two different aspects need to be investigated to achieve these proposed results. On
one hand, there is the classification or the way to aggregate catchments based on a

EuroLandscape CCM, April 2000

20

Francesca Bertolo

set of characteristics. On the other hand, there is the necessity to understand what is
representative.
Classification is defined as the ordering of objects into groups or sets on the basis
of their similarities or relationships; the subsequent allocation or assignment of
objects into the pre-established classification system is called identification
(Sokal, 1974). A major contribution of a land classification to resource studies on a
regional scale is that spatial patterns become evident. The relationship between
terrain features and physical and biological processes are almost always
characterized by spatial patterns. A land classification system may be built either on
taxonomic site classification or regionalisation. The taxonomic approach seeks to
delineate zones by grouping sites with similar properties, while a regionalisation
approach subdivides land hierarchically into land units on the basis of presumptive
rules. The inherent differences between these two approaches can be simply reduced
to whether an inductive process or a deductive process is applied. In the
regionalisation approach, land classification systems have been set-up based on
logical rules and the methodology subsequently tailored to fit principles. Analytical
regionalisation is basically a mapping procedure and often implies knowledge of the
causes of similarities and differences between the objects studied.
For long time hydrologists have been striving for the perfect model that can be
applied to all catchments. This has apparently led to inefficiencies and overparameterisations in rainfall-runoff modelling, giving rise to some questioning about
the value of complex, spatially distributed models for hydrologic predictions
(Grayson et al., 1992).
As an alternative, theories that would enable to transfer hydrologic information and
understanding from one catchment to another have been searched. Flood estimation
in ungauged catchments, for example, is quite often carried out by a regional
transfer of rainfall-runoff and/or flood frequency data from one catchment to
another. Success in such regionalisations depends crucially upon defining the
conditions under which two catchments may be considered to be hydrologically
similar. Definitions of similarity used in the past have been based on
physioclimatic characterizations without explicit recognition of the environmental
controls on runoff generation. An alternative point of view suggest that, if
catchments can be classified in terms of the similarity of their runoff generation
responses, specialised models can be developed to suit the actual mechanisms of
runoff generation that operate on a given catchment. Similarity considerations can
help in guiding the design of field experiments and the interpretation of data from
such field studies (Larsen et al., 1994).
The theoretical concepts of similarity were first proposed by Sivapalan et al. (1987)
to provide a framework for a comparison of runoff generation responses from
catchments with different characteristics and of different sizes. By recasting the
equations of a rainfall-runoff model into dimensionless forms, they were able to
identify five dimensionless similarity parameters as potential measure of hydrologic
JRC SAI EGEO

Catchment Delineation and Characterisation

21

similarity. These parameters are based on field-measured soil and topographic


properties such as the saturated hydraulic conductivity at the soil surface, the
thickness of the capillarity fringe above the water table, mean pore size, saturated
and residual soil moisture content and a topographic index (Larsen et al., 1994).
In this general framework the following studies are interesting, because they
describe what kind of work has been carried out until today and how close the
international research is to the aims of the EuroLandscape project.
Huang and Ferng (1990) used a taxonomic approach to classify land units
objectively into homogeneous zones for water quality management purposes. The
framework for the classification is divided into two parts: catchment classification
and land process classification. In order to reveal interrelations among the spatial
patterns of both land and water characteristics, as well as human factors, catchment
classification is subdivided in three parts: drainage basin morphometry, assimilative
capacity of stream and water quality management. The purpose of the catchment
classification with particular reference to basin morphometry is to describe the
spatial pattern of terrain characteristics and to analyse their relevance for water
quality management. The classification based on the assimilative capacity
emphasises the potential influence of land characteristics on the resilience and
stability of surface water. The water quality management classification represents
the final product of the catchment classification scheme, which not only includes
factors of land characteristics related to water quality, but also incorporates the
relative importance of each catchment for human society, specifically the designated
stream classes and water quality standards.
In practice, variables assumed to affect stream quality and quantity were measured
in each of 78 catchments of the study area, the Tanshui River Basin in northern
Taiwan. Variables were selected only if they could be readily measured or derived
from published maps. The selected variables are: location, area, basin configuration,
elevation, slope, surface roughness, parent material, texture, drainage, stream
number, stream length, stream frequency, stream density, bifurcation ratio, streambed slope, water quality standards, precipitation, evapotranspiration, land use of the
catchment and in particular of the riparian zones. The framework of the
classification proposed in this research is of a multivariate nature consisting in a
three-step procedure utilising both clustering and ordination techniques for each of
the three catchment classifications described previously, each time with a specific
subset of variables.
Civco et al. (1995) described the effective catchment characterization project as a
procedure that typically follows four basic steps. The first step is to establish a main
objective; the spatial extent and degree of characterization depend on that objective.
The second step is to perform a general search to ascertain what data are available,
followed by acquiring general data. The third step is to select the catchments using a
multi-step process in which the first selection is based on general data, and
subsequent selections are based on more specific data. The fourth step is the
EuroLandscape CCM, April 2000

22

Francesca Bertolo

catchment characterization itself. They also highlighted that GIS is an ideal tool in
this kind of applications: scientists must work with data such land use and land
cover, geology, soils, hydrography and topography, and this usually involves
integrating multiple, diverse data sources and types.
Lent et al. (1998) described another example of multivariate analysis for a
classification of small order catchments of the Quabbin Reservoir basin,
Massachusetts (USA), related to water quality. Effective management of water
quality in a large drainage basin that is comprised of numerous small catchments
may require aggregation of them into a small number of sub-basins on the basis of
landscapes features that affect the quantity and quality of water. Meaningful
catchment aggregation is predicated on identifying the hydrologic and
biogeochemical processes that are most important in controlling stream water
chemistry and in understanding how those processes are distributed among
catchments throughout the large drainage basin.
Based on eleven stream water chemical constituents a first classification of 15
gauged catchments was conducted. With principal components analysis (PCA) the
median water quality concentrations were reduced to three principal components,
then the extent to which these 15 stations were related to a principal component was
measured with the so-called stations scores of the three principal components. PC1
and PC2 scores varied geographically from west to east, indicating that there were
systematic spatial variations in water chemistry. This grouping based on geographic
location formed the basis for a preliminary classification of catchments and three
groups were defined: the west, central, and east sub-basin. In a second stage analysis
of variance and multiple comparisons of means were used to test for significant
differences among sub-basins in the magnitude of the stream water variables and in
24 landscape attributes (slopes, bedrock, soils, land-use, roads lengths and detailed
wetland data). These showed that differences in stream water chemistry were
consistent with variations in landscape attributes. Then multiple regression analysis
was used to develop relations between stream water chemistry, defined by the
stations scores from the PCA, to selected landscape attributes. The analysis resulted
in three significant equations, one for each of the three principal components.

6. Final Considerations
Several topics have been investigated in this review. Not in all cases it was possible
to find a clear answer to the questions formulated at the beginning. A few points
could, however, be highlighted that will be useful in the Catchment Characterisation
and Modelling activity. The most pertinent conclusions are the following:
(1) There is a need for some method to quantify errors. It is clear that a raster DEM
is a partial representation of the landscape and that slope is the derived parameter,
which suffers most from that partiality. As a consequence, the delineation of
JRC SAI EGEO

Catchment Delineation and Characterisation

23

catchment boundaries suffers from a high degree of uncertainty, independently from


the quality and the geometric resolution of the underlying DEM and independently
from the method used to define drainage pathways. There is not clear indication of
how to quantify this error.
(2) Considering the necessary spatial resolution of the DEM and the extended area
to be covered, there are only few DEMs available. The USGS GTOPO30 DEM has
been widely used and it proved to be of good quality in the European part. However,
its resolution is too coarse to achieve the targeted final map scale of 1:250.000.
Buying commercially available DEMs for the whole area is not affordable at this
moment in time. A possible alternative could be the DEM resulting from the STRM
mission. However, neither the distribution modalities nor the price policy have yet
been defined.
(3) As can be seen from the large number of publications, there has been a lot of
interest in river network delineation in the last years. This aspect is much more
relevant than the delineation of the catchment boundaries. Technically it is a
relatively simple problem to define catchment boundaries once the drainage
channels are identified. For this reason, the report gives most attention to the
techniques to model the flow paths.
Several methods have been described and the different aspects of the problem have
been evaluated: convergent flow, divergent flow, non-point diffusion, undefined
flow direction, flat area problem, etc. From this point of view the choice of an
appropriate algorithm is difficult. However, many of these algorithms are not
available in ready to use software packages or require too large computational
efforts and, therefore, are not adequate to handle large DEMs.
(4) The correct definition of source areas is still an open problem. Considerable
discussions have been conducted on this topic, but generally there is no solution
applicable to DEMs of medium resolution. The problem is that the theories on the
channel initialisation are generally based on studies conducted on the field scale and
working with DEMs of high and very high resolution. The source area size is
generally smaller than the grid size of a medium or low resolution DEM. In general,
a quick solution is to select an arbitrary size of the source area and to adjust it
iteratively in order to approximate the extent of mapped river networks, which serve
as a reference. At the European scale it is difficult to obtain a complete and
comprehensive river network to use as a reference. Therefore, there is a clear need
for some technique to bridge the gap between the theories and the available grid
resolution.
(5) The final coding system seems to be technically complicated by the large
number of catchments involved, but there exist many examples at regional level that
can be helpful. In the framework of this project it has been decided to aim at a
minimum catchment area of around 500 to 1000 km2. Considering that the European
area to be mapped extends from the Mediterranean to northern Scandinavia and
EuroLandscape CCM, April 2000

24

Francesca Bertolo

from Portugal to the Urals and corresponds to a surface of about 10.2 106 km2
(Stanners and Bourdeau, 1995: 19), this will result in an approximate number of
10,000 to 20,000 catchments.

7. References
A.

Cited in the Text

Band, L. E. (1986): Topographic partition of watersheds with Digital Elevation Models - Water
Resources Research, 22, pp.15-24.
Bolstard, P. V.; Stowe, T. (1994): An evaluation of DEM accuracy: elevation, slope and aspect Photogr. Eng. and Rem. Sens., 60, pp.1327-1332.
Chorowicz, J.; Ichoku, C.; Riazanoff, S.; Kim, Y.; Cervelle, B. (1992): A combined algorithm for
automated drainage network extraction - Water Resources Research, 28, pp.1293-1302.
Civco, D. L.; Garcia, A. R.; Warner, G. S. (1995): Key steps to effective watershed
characterisation - GIS World, 8, pp.62-67.
Costa-Cabral, M. C.; Burges, S. J. (1994): Digital elevation model networks (DEMON): a Model
of flow over hillslopes for computation of contributing and dispersal areas - Water Resources
Research, 30, pp.1681-1692.
DaRos, D.; Borga, M. (1997): Use of digital elevation model data for the derivation of the
geomorfological istantaneous unit hydrograph - Hydrological Processes, 11, pp.13-33.
Desmet, P. J. J. (1997): Effects of interpolation errors on the analysis of DEMs - Earth Surface
Processes and Landforms, 22, pp.563-580.
Fairfield, J.; Laymarie, P. (1991): Drainage Networks from Grid Digital Elevation Models - Water
Resources Research, 27, pp.709-717.
Flavin, R. W.; Andrews, A. J.; Kronvang, B.; Mller-Wohlfeil, D.; Demuth, S.; Birkenmayer,
A. (1998): ERICA, European Rivers and Catchments (European Environment Agency)
Copenhagen.
Freeman, T. G. (1991): Calculating catchment area with divergent flow based on a regular grid Computers & Geosciences, 17, pp.413-422.
Garbrecht, J.; Martz, L. (1994): Grid size dependency of Parameters extracted from digital
elevation Models - Computers & Geosciences, 20, pp.85-87.
Garbrecht, J.; Martz, L. W. (1997): The assignment of drainage direction over flat surfaces in
raster digital elevation models - Journal of Hydrology, 193, pp.204-213.
Giles, P. T.; Franklin, S. E. (1996): Comparison of derivative topographic surfaces of a DEM
generated from stereoscopic SPOT images with field measurements - Photogrammetric
Engineering & Remote Sensing, 62, pp.1165-1171.
Grayson, R.B.; Moore, I.D.; McMahon, T.A. (1992): Physically based hydrologic modelling, 2, Is
the concept realistic? - Water Resources Research, 28, pp.2659-2666.
Haralick, R.M. (1983): Ridge and valley on digital images Computer Vision, Graphics, and Image
processing, 22, pp.28-38.
Helmlinger, K. R.; Kumar, P.; Foufoula-Georgiou, E. (1993): On the Use of Digital Elevation
Model Data for Hortonian and Fractal Analyses of Channel Networks - Water Resources
Research, 29, pp.2599-2613.
Hjelmfelt, A.T. (1988): Fractals and the river-length catchment-ratio - Water Resources Bulletin, 24,
pp.455-459.
Huang, S.; Ferng, J. (1990): Applied land classification for surface water quality management: I.
Watershed classification - Journal of Environmental Management, 31, pp.107-126.

JRC SAI EGEO

Catchment Delineation and Characterisation

25

Hussein, M.; Schwartz, F. W. (1997): Application of terrain analysis and geographic information
systems to the construction of hydrologic databases: A case study - Environmental &
Engineering Geoscience, 3, pp.549-562.
Hutchinson, M. F. (1988): Calculation of hydrologically sound digital elevation models - Proc.
Third International Symposium on Spatial Data Handling.
Sydney, Australia
Sydney, Australia, pp. 117-133.
Hutchinson, M. F. (1989): A new procedure for gridding elevation and stream line data with
automatic removal of spurious pits - Journal of Hydrology, 106, pp.211-232.
Ijjasz-Vasquez, E. J.; Bras, R. L. (1995): Scaling regimes of local slope versus contributing area in
digital elevation models - Geomorphology, 12, pp.299-311.
Isaacson, D. L.; Ripple, W. J. (1990): Comparison of 7.5-Minute and 1-Degree Digital Elevation
Models - Photogrammetric Engineering & Remote Sensing, 56, pp.1523-1527.
Jenson, S. K.; Domingue, J. O. (1988): Extracting topographic structure from digital elevation data
for geographic information system analysis - Photogrammetric Engineering and Remote
Sensing, 54, pp.1593-1600.
Jones, N. L.; Wright, S. G.; Maidment, D. R. (1990): Watershed delineation with triangle-based
terrain models - Journal of Hydraulic Engineering, 116, pp.1232-1251.
La Barbera, P.; Rosso, R. (1989): On the fractal dimension of stream networks - Water Resources
Research, 25, pp.735-741.
Lammers, R. B.; Band, L. E. (1990): Automatic object representation of drainage basins Computers & Geosciences, 16, pp.787-810.
Larsen, J. E.; Sivapalan, M.; Coles, N. A.; Linnet, P. E. (1994): Similarity analysis of runoff
generation processes in real-world catchments - Water Resources Research, 30, pp.1641-1652.
Lent, R. M.; Waldron, M. C.; Rader, J. C. (1998): Multivariate classification of small order
watersheds in the Quabbin Reservoir Basin, Massachusetts - Journal of the American Water
Resources Association, 34, pp.439-450.
Mackay, S. D.; Band, L. E. (1998): Extraction and representation of nested catchment areas from
digital elevation models in lake-dominated topography - Water Resources Research, 34,
pp.897-901.
Mandelbrot, B.B. (1982): The Fractal Geometry of Nature. (W.H. Freeman) New York.
Mark, D. M. (1984): Automatic detection of drainage networks from digital elevation models Cartographica, 21, pp.168-178.
Martz, L. W.; Garbrecht, J. (1995): Automated recognition of valley line and drainage networks
from grid digital elevation models: a review and a new method Comment - Journal of
Hydrology, 167, pp.393-396.
Martz, L. W.; Garbrecht, J. (1998): The treatment of flat areas and depressions in automated
drainage analysis of raster digital elevation models - Hydrological Processes, 12, pp.843-855.
Meisels, A.; Raizman, S.; Karnieli, A. (1995): Skeletonizing a DEM into a drainage network Computers & Geosciences, 21, pp.187-196.
Miller, D. R.; Morrice, J.G. (1996): Assesing uncertainty in catchment boundary delimitation
Proc.Third Int. Conference/Workshop on Integrating GIS and Environmental Modeling. Santa
Fe, NM, USA, January 21-26, 1996. National Center for Geographic Information and
Analysis, Santa Barbara, CA, USA. http://www.ncgia.ucsb.edu/conf/SANTA_FE_CDROM/main.html.
Montgomery, D. R.; Dietrich, W. E. (1988): Where do channels begin? - Nature, 336, pp.232-234.
Montgomery, D. R.; Dietrich, W. E. (1989): Source areas, drainage density, and channel initiation Water Resources Research, 25, pp.1907-1918.
Montgomery, D. R.; Dietrich, W. E. (1992): Channel initiation and the problem of landscape scale Science, 255, pp.826-830.
Montgomery, D. R.; Foufoula-Georgiou, E. (1993): Channel Network Source Representation
Using Digital Elevation Models - Water Resources Research, 29, pp.3925-3934.
Moore, I. D.; Grayson, R. B. (1991): Terrain-based catchment partitioning and runoff prediction
using vector elevation data - Water Resources Research, 27, pp.1177-1191.

EuroLandscape CCM, April 2000

26

Francesca Bertolo

Moussa, R.; Bocquillon, C. (1996): Fractal analyses of tree-like channel networks from digital
elevation model data - Journal of Hydrology, 187, pp.157-172.
Nelson, E. J.; Jones, N. L. (1995): Reducing elevation roundoff errors in digital elevation models Journal of Hydrology, 169, pp.37-49.
O'Callaghan, J. F.; Mark, D. M. (1984): The extraction of drainage networks from digital elevation
data - Computer Vision, Graphics, and Image Processing, 28, pp.323-344.
Palacios-Velez, O. L.; Cuevas-Renaud, B. (1986): Automated river-course, ridge and basin
delineation from digital elevation data .- Journal of Hydrology, 86, pp.299-314.
Peucker, T. K.; Douglas, D. H. (1975): Detection of surface-specific points by local parallel
processing of discrete terrain elevation data - Computer Graphics and Image Processing, 4,
pp.375-387.
Riazanoff, S.; Julien, P.; Cervelle, B.; Chorowicz, J. (1992): Extraction et analyse automatiques
d'un rseau hirarchis de talwegs. Application un modle numrique de terrain driv d'un
couple stroscopique SPOT - Int. J. Rem. Sens., 13, pp.337-364.
Roth, G.; La Barbera, P.; Greco, M. (1996): On the description of the basin effective drainage
structure - Journal of Hydrology, 187, pp.119-135.
Sabbagh, G. J.; Storm, D. R.; Haan, C. T. (1994): Digital terrain model to estimate watershed
characteristics for hydrology and water quality modelling - Proc. American Society
Agricultural Engineers Summer Meeting. June 19-22 1994. Kansas City, Missouri, USA, pp.
Paper No.94-2034.
Seaber, P. R.; Kapinos, F. P.; Knapp, G. L. (1987): Hydrologic Unit Maps - United States
Geological Survey Water-Supply Paper,, pp.63.
Sekulin, A. E.; Bullock, A.; Gustard, A. (1992): Rapid Calculation of Catchment Boundaries Using
an Automated River Network Overlay Technique - Water Resources Research, 28, pp.21012109.
Sivapaln, M.; Beven, K.; Wood, E.F. (1987): On hydrologic similarity, 2, a scaled model of storm
runoff production - Water Resources Research, 23, pp.2266-2278.
Skidmore, A. K. (1990): Terrain position as mapped from a gridded digital elevation model - Int. J.
Geographical Information Systems, 4, pp.33-49.
Smith, T. R.; Zhan, C.; Gao, P. (1990): A knowledge-based, two-step procedure for extracting
channel networks from noisy DEM data - Computers & Geosciences, 16, pp.777-786.
Sokal, R. R. (1974): Classification: purposes, principles, progress, prospects - Science, 185, pp.11151123.
Speight; J.G. (1968): Parametric Description of Land Form - Proc. Land evaluation, CSIRO
Symposium in cooperation with UNESCO. Canberra, Australia, pp. 239-250.
Stanners, D.; Bourdeau, Ph. (1995): Europes Environment. The Dobrs Assessment.- European
Environmental Agency, Copenhagen, 676 p.
Tarboton, D. G. (1997): A new method for the determination of flow directions and upslope areas in
grid digital elevation models - Water Resources Research, 33, pp.309-319.
Tarboton, D. G.; Bras, R. L.; Rodriguez-Iturbe I. (1989): Scaling and elevation in river networks Water Resources Research, 25, pp.2037-2051.
Tarboton, D. G.; Bras, R. L.; Rodriguez-Iturbe, I. (1992): A physical basis for drainage density Geomorphology, 5, pp.59-76.
Thieken, A. H.; Lucke, A.; Diekkruger, B.; Richter, O. (1999): Scaling input data by GIS for
hydrological modelling - Hydrological Processes, 13, pp.611-630.
Tribe, A. (1992): Auromated recognition of valley lines and drainage networks from grid digital
elevation models: a review and a new method - Journal of Hydrology, 139, pp.263-293.
Verdin, K. L.; Jenson, S. K. (1996): Development of continental scale DEMs and extraction of
hydrographic features Verdin, K. L.; Verdin, J. P. (1999): A topological system for delineation and codification of the
Earth's river basins - Journal of Hydrology, 218, pp.1-12.
Verheust, L.; Johnson, G. (1998): The watershed database for sub-equatorial Africa, structure and
user interface. ALCOM Working Paper No 16 (ALCOM/FAO) Harare, 21 p.

JRC SAI EGEO

Catchment Delineation and Characterisation

27

Walker, J. P.; Willgoose, G. R. (1999): On the effect of digital elevation model accuracy on
hydrology and geomorphology - Water Resources Research, 35, pp.2259-2268.
Wang, X.; Yin, Z.-Y. (1998): A comparison of drainage networks derived from digital elevation
models at two scales - Journal of Hydrology, 210, pp.221-241.
White, J. D.; Running, S. W. (1994): Testing scale dependant assumptions in regional ecosystem
simulations - Journal of Vegetation Sciences, 5, pp.687-702.
Wolock, D. M.; Price, C. V. (1994): Effects of digital elevation model map scale and data resolution
on a topography-based watershed model - Water Resources Research, 30, pp.3041-3052.
Yoeli, P. (1984): Computer-assisted determination of the valley and ridge lines of digital terrain
models Int. Yearbook of Cartography, 24, pp.197-205.
Zhang, W.; Montgomery, D. R. (1994): Digital elevation model grid size, landscape representation,
and hydrologic simulations - Water Resources Research, 30, pp.1019-1028.

B.

Further Reading

Arnell, N. W. (1999): A simple water balance model for the simulation of streanflow over a large
geographic domain - Journal of Hydrology, 217, pp.314-335.
Bailey, R. G.; Pfister, R. D.; Henderson, J. A. (1978): Nature of land and resources classification A review - Journal of Forestry, 76, pp.650-655.
Becker, A.; Braun, P. (1999): Disaggregation, aggregation and spatial scaling in hydrological
modelling - Journal of Hydrology, 217, pp.239-252.
Benosky, C. P.; Merry, C. J. (1995): Automatic extraction of watershed characteristics using spatial
analysis techniques with application to groundwater mapping - Journal of Hydrology, 173,
pp.145-163.
Bischetti, G. B.; Gandolfi, C.; Whelan M.J. (1998): The definition of stream channel head location
using digital elevation data - Proc. HeadWater'98: Hydrology, Water resources and Ecology
in Headwaters. Meran/Merano, Italy, pp. 545-552.
Blaszczynski, J. S. (1997): Landform Characterization with Geographic Information Systems Photogrammetric Engineering & Remote Sensing, 63, pp.183-191.
Carpenter, T. M.; Sperfslage, J. A.; Georgakakos, K. P.; Sweeney, T.; Fread, D. L. (1999):
National threshold runoff estimation utilizing GIS in support of operational flash flood
warning systems - Journal of Hydrology, 224, pp.21-44.
Cialella, A. T.; Dubayah, R.; Lawrence, W.; Levine, E. (1997): Predicting Soil Drainage Class
Using Remotely Sensed and Digital Elevation Data - Photogrammetric Engineering & Remote
Sensing, 63, pp.171-178.
Dawes, W. R.; Short, D. (1994): The significance of topology for modeling the surface hydrology of
fluvial landscapes - Water Resources Research, 30, pp.1045-1055.
DeVantier, B. A.; Feldman, A. D. (1993): Review of GIS applications in hydrologic modelling Journal of Water Resources Planning and Management, 119, pp.246-261.
Dietrich, W. E.; Wilson, C. J.; Montgomery, D. R.; McKean, J. (1993): Analysis of erosion
thresholds, channel networks, and landscape morphology using a digital terrain model Journal of Geology, 101, pp.259-278.
Dikau, R. (1989): The application of a digital relief model to landform analysis in geomorphology In: Raper, J. (ed): Three dimensional applications in GIS. (Taylor & Francis), pp. 51-77.
Eash, D. A. (1994): A geographic information system procedure to quantify drainage-basin
characteristics - Water Resources Bulletin, 30, pp.1-8.
Engman, E. T. (1995): The use of remote sensing data in watershed research - Journal of Soil &
Water Conservation, 50, pp.438-440.
Fels, J.E.; Matson, K.C. (1996): A cognitively-based approach for hydrogeomorphic land
classification using digital elevation models - Proc.Third Int. Conference/Workshop on
Integrating GIS and Environmental Modeling. Santa Fe, NM, USA, January 21-26, 1996.
National Center for Geographic Information and Analysis, Santa Barbara, CA, USA.
http://www.ncgia.ucsb.edu/conf/SANTA_FE_CD-ROM/main.html.
EuroLandscape CCM, April 2000

28

Francesca Bertolo

Fernandez, B. L.; Pizarro, G. P. (1996): Statistical estimation of runoff characteristics of


watersheds in central Chile - Hydrological Sciences Journal, 41, pp.735-749.
Florinsky, I. V. (1998): Combined analysis of digital terrain models and remotely sensed data in
landscape investigations - Progress in Physical Geography, 22, pp.33-60.
Gallant, J.; Hutchinson, M.F. (1996): Towards an understanding of landscape scale and structure Proc.Third Int. Conference/Workshop on Integrating GIS and Environmental Modeling. Santa
Fe, NM, USA, January 21-26, 1996. National Center for Geographic Information and
Analysis, Santa Barbara, CA, USA. http://www.ncgia.ucsb.edu/conf/SANTA_FE_CDROM/main.html.
Gallant, J.; Wilson, J. P. (1996): Tapes-G: A grid-based terrain analysis program for the
environmental sciences - Computers & Geosciences, 22, pp.713-722.
Gandolfi, C.; Bischetti, G. B. (1997): Influence of the drainage network identification method on
geomorphological properties and hydrological response - Hydrological Processes, 11, pp.353375.
Garbrecht, J.; Martz, L. W. (1993): Comment on "A Combined Algorithm for Automated
Drainage Network Extraction" by Jean Chorowicz et al. - Water Resources Research, 29,
pp.535-536.
Garbrecht. J.; Martz, L. W. (1997): Automated Channel ordering and Node indexing for Raster
Channel Networks - Computers & Geosciences, 23, pp.961-966.
Graham, S. T.; Famiglietti, J. S.; Maidment, D. R. (1999): Five-minute, 1/2, and 1 degree data sets
of continental watersheds and river networks for use in regional and global hydrologic and
climate system modeling studies - Water Resources Research, 35, pp.583-587.
Goodrich, D. C.; Woolhiser, D. A. (1991): Catchment Hydrology - Reviews of Geophysics, 29,
pp.202-209.
Gyasi-Agyei, Y.; Willgoose, G.; De Troch, F. P. (1995): Effects of vertical resolution and map
scale of digital elevation models on geomorphological parameters used in hydrology Hydrological Processes, 9, pp.363-382.
Gyasi-Agyei, Y.; De Troch, F. P.; Troch. P. A. (1996): A dynamic hillslope response model in a
geomorphology based rainfall-runoff model - Journal of Hydrology, 178, pp.1-18.
Hagemann, S.; Dumenil, L. (1998): A parametrization of the lateral waterflow for the global scale Climate Dynamics, 14, pp.17-31.
Hornberger, G. M.; Boyer, E. W. (1995): Recent advances in watershed modelling - Reviews of
Geophysics, 33, pp.949-957.
Howard, A. D. (1994): A detachment-limited model of drainage basin evolution - Water Resources
Research, 30, pp.2261-2285.
Hunter, G. J.; Goodchild, M. F. (1995): Dealing with Error in Spatial Databases: a Simple Case
Study - Photogrammetric Engineering & Remote Sensing, 61, pp.529-537.
Hutchinson, M. F.; Dowling, T. I. (1991): A continental hydrological assessment of a new gridbased digital elevation model of Australia - Hydrological Processes, 5, pp.45-58.
Imhof, J. G.; Fitzgibbon, J.; Annable, W. K. (1996): A hierarchical evaluation system for
characterizing watershed ecosystem for fish habitat - Canadian Journal of Fisheries and
Aquatic Sciences, 53, pp.312-326.
Irvin, B. J.; Ventura, S. J.; Slater, B. K. (1997): Fuzzy and isodata classification of landform
elements from digital terrain data in Pleasant Valley, Wisconsin - Geoderma, 77, pp.137-154.
Kirkby, M. J.; Imeson, A. C.; Bergkamp, G.; Cammeraat, L. H. (1996): Scaling up processes and
models from the field plot to the watershed and regional areas - Journal of Soil & Water
Conservation, 51, pp.391-396.
Kite, G. W.; Kouwen, N. (1992): Watershed Modeling Using Land Classifications - Water
Resources Research, 28, pp.3193-3200.
Kite, G. W.; Dalton, A.; Dion, K. (1994): Simulation of streamflow in a macroscale watershed
using general circulation model data - Water Resources Research, 30, pp.1547-1559.

JRC SAI EGEO

Catchment Delineation and Characterisation

29

Lee, J.; Snyder, P. K.; Fisher. P. F. (1992): Modeling the Effects of Data Errors on Feature
Extraction from Digital Elevation Models - Photogrammetric Engineering & Remote Sensing,
58, pp.1461-1467.
Lotspeich, F. B. (1980): Watersheds as the basic ecosystem: this conceptual framework provides a
basis for a natural classification system - Water Resources Bulletin, 16, pp.581-586.
Lunetta, R. S.; Cosentino, B. L.; Montgomery, D. R.; Beamer, E. M.; Beechie, T. J. (1997): GISbased evaluation of salmon habitat in the Pacific Northwest - Photogrammetric Engineering
and Remote Sensing, 63, pp.1219-1229.
MacMillan, R. A.; Furley, P. A.; Healey, R. G. (1993): Using hydrological models and geographic
information systems to assist with the management of surface water in agricultural landscapes
- In: Haynes-Young, R.; Green, D. R. und Cousins, S. (eds): Landscape ecology and
geographic information systems. (Taylor & Francis) London, New York, Philadelphia, pp.
181-209.
Marks, D.; Dozier, J.; Frew, J. (1984): Automated basin delineation from digital elevation data Geo-Processing, 2, pp.299-311.
Matson, K. C.; Stallings, C.; Jennings, G. D.; Mclaughlin, R. A.; Coffey, S. W. (1995):
Watershed scale hydrologic modelling in a GIS environment - Proc. GIS/LIS '95, Annual
Conference and Expo. Nashville, TN, USA, pp. 713-722.
McCormack, J. E.; Gahegan, M. N.; Roberts, S. A.; Hogg, J.; Hoyle, B. S. (1993): Feature-based
derivation of drainage networks - International Journal of Geographical Information Systems,
7, pp.263-279.
McDonnell, R. A. (1996): Including the spatial dimension: using geographical information systems
in hydrology - Progress in Physical Geography, 20, pp.159-177.
Morris, E. M.; Blyth, K.; Clarke, R. T. (1980): Watershed and river channel characteristics, and
their use in a mathematical model to predict flood hydrographs.- In: Fraysse G.(ed): Remote
Sensing Application in Agriculture and Hydrology. (A. A. Balkema) Rotterdam, pp. 431-446.
Moore, I. D.; Grayson, R. B.; Ladson, A. R. (1991): Digital terrain modelling: a review of
hydrological, geomorphological, and biological applications - Hydrological Processes, 5,
pp.3-30.
Moussa, R. (1997): Geomorphological transfer function calculated from digital elevation models for
distributed hydrological modelling - Hydrological Processes, 11, pp.429-449.
O'Donnell, G.; Nijssen, B.; Lettenmaier, D. P. (1999): A simple algorithm for generating
streamflow networks for grid-based, macroscale hydrological models - Hydrological
Processes, 13, pp.1269-1275.
Paracchini, M. L.; Folving, S.; Bertolo, F. (1999): Identification and classification of European
headwaters - Proc. 21st Session of the European Forestry Working Party on the Management
of Mountain Watersheds.
Prosser, I. P. (1996): Tresholds of channel initiation in historical and holocene times, Southeastern
Australia.- In: Anderson, M. G. und Brooks, S. M.(eds): Advances in Hillslope Processes.
(John Wiley & Sons Ltd.), pp. 687-708.
Quinn, P. F.; Beven, K. J.; Lamb, R. (1995): The ln(a/tan) index: how to calculate it and how to
use it within the TOPMODEL framework - Hydrological Processes, 9, pp.161-182.
Ravichandran, S.; Ramanibai, R.; Pundarikanthan, N. V. (1996): Ecoregions for describing
water quality patterns in Tamiraparani basin, South India - Journal of Hydrology, 178, pp.257276.
Riggs, H. C. (1973): Regional analyses of streamflow characteristics - Techniques of WaterResources Investigations Book 4. (USGS), 15 p.
Rodda, J. C.; Pieyns, S. A.; Sehmi, N. S.; Matthews, G. (1993): Towards a world hydrological
observing system - Hydrological Sciences Journal, 38, pp.373-378.
Rodriguez-Iturb, e. I.; Ijjasz-Vasquez, E. J.; Bras, R. L.; Tarboton, D. G. (1992): Power law
distribution of discharge mass and energy in river basins - Water Resources Research, 28,
pp.1089-1093.

EuroLandscape CCM, April 2000

30

Francesca Bertolo

Rosso, R.; Bacchi, B.; La Barbera, P. (1991): Fractal relation of mainstream length to catchment
area in river networks - Water Resources Research, 27, pp.381-387.
Sefton, C. E. M.; Howarth, S. M. (1998): Relationship between dynamic response characteristics
and physical descriptors of catchments in England and Wales - Journal of Hydrology, 211,
pp.1-16.
Smith, M. B.; Brilly, M. (1992): Automated grid element ordering for GIS-based overland flow
modelling - Photogrammetric Engineering and Remote Sensing, 58, pp.579-585.
Snell, J. D.; Sivapalan, M. (1994): On geomorphological dispersion in natural catchments and the
geomorphological unit hydrograph - Water Resources Research, 30, pp.2311-2323.
Soille, P. J.; Ansoult, M. M. (1990): Automated basin delineation from digital elevation models
using mathematicaal morphology - Signal Processing, 20, pp.171-182.
Soille, P. J.; Gratin, C. (1994): An efficient algorithm for drainage network extraction on DEMs Journal of Visual Communication and image Representation, 5, pp.181-189.
Steinwand, D. R.; Hutchinson, J. A.; Snyder, J. P. (1995): Map projections for global and
continental data sets and an analysis of pixel distortion caused by reprojection Photogrammetric Engineering and Remote Sensing, 61, pp.1487-1497.
Tarboton, D. G.; Bras, R. L.; Rodriguez-Iturbe, I. (1991): On the extraction of channel networks
from digital elevation data - Hydrological Processes, 5, pp.81-100.
Thorsteinson, L. K.; Taylor, D. L. (1997): A watershed approach to ecosystem monitoring in
Denali National Park and Preserve, Alaska - Journal of the American Water Resources
Association, 33, pp.795-810.
Walsh, S. J.; Butler, D. R.; Malanson, G. P. (1998): An overview of scale, pattern, process
relationships in geomorphology: a remote sensing and GIS perspective - Geomorphology, 21,
pp.183-205.
Wilson, J. P.; Gallant, J. C. (1996): EROS: A grid-based program for estimating spatiallydistributed erosion indices - Computers and Geosciences, 22, pp.707-712.
Wolock, D. M. (1995): Effects of subbasin size on topographic characteristics and simulated flow Water Resources Research, 31, pp.1989-1997.
Wood, E. F.; Sivapalan, M.; Beven, K.; Band, L. (1988): Effects of spatial variability and scale
with implications to hydrologic modelling - Journal of Hydrology, 102, pp.29-47.

JRC SAI EGEO

Catchment Delineation and Characterisation. A Review.


by Francesca Bertolo
EuroLandscape Catchment Characterisation and Modelling (CCM), Report No. 1

2000, 36 p.
JRC EUR 19563 EN
European Commission, DG Joint Research Centre (JRC), Space Applications Institute
(SAI), Environment and Geo-Information (EGEO) Unit, 21020 Ispra (Va), Italy

In the frame of the EuroLandscape project, the Space Applications Institute (SAI) of the
Joint Research Centre (JRC) of the European Commission is working towards the creation
of a pan-European database of catchment boundaries and river networks. Together with the
physical and socio-economic characterisation of the mapped catchments, this database will
serve for the analysis of data on water quality and quantity collected by the European
Environmental Agency (EEA) as well as for the modelling of hydrological and landscape
processes throughout Europe.
This report reviews the possibilities and limitations to delineate catchments and drainage
networks from digital elevation models of different grid-cell sizes, with specific
consideration of the large area to be covered by the Catchment Characterisation and
Modelling (CCM) activity of EuroLandscape. It further discusses issues related to the
ordering of catchments and drainage networks as well as issues related to their
characterisation.

Вам также может понравиться