A GIS-BASED TERRAIN ANALYSIS APPROACH FOR WETLAND
INVENTORY IN THE SEMI-ARID HEADWATERS OF THE
UMZIMVUBU BASIN, SOUTH AFRICA
H. Dahlkea, J. Helmschrota and T. Behrensb
Department of Geoinformatics, Geohydrology and Modelling, FSU Jena, Loebdergraben 32,
D-07743 Jena, Germany, email: Helen.Dahlke@uni-jena.de
Department of Physical Geography and Soil Science, FSU Jena
The presented study focuses on the development of a terrain-based method to identify and delineate
different types of palustrine wetlands in semi-arid headwater catchments of the Drakensberg Mountains,
South Africa. The vulnerable wetland ecosystems are facing an increasing pressure due to noticeable land
use changes in this region. Detailed information due to the distribution and the extent of different wetland
types within these catchments are necessary in order to analyse, model and assess impacts caused by
afforestation. A comprehensive system analysis based on intensive field work indicates a close relationship
between the hydrodynamics of the wetlands and their terrain characteristics. Therefore a hydro-geomorphic
classification approach based on the combination of several classification schemes has been developed to
differentiate wetland types in terms of their terrain position, morphometry as well as their
surface/subsurface hydrological regime. This approach was realized in a GIS-based analysis system to
delineate different wetland types by the variation of specific geomorphometric and geomorphographic
parameters. The delineation of these parameters is based on a high quality SRTM-DEM of 25x25 m² grid
size. It could be shown, that a combination of different types of curvatures, the compound topographic
index and parameters indicating hillslope position provides excellent results for the delineation of different
wetland types. The delineated wetland distribution map has been validated with land use data, field survey
and Landsat TM/ETM data. An overall prediction accuracy of 94 % approved the potential of the
developed method for a variety of environmental research objectives.
Keywords: wetlands, terrain-analysis, Shuttle Radar Topography Mission (SRTM), DEM, South Africa
The landscape of the semi-arid Eastern Cape Province, South Africa is characterized by the occurrence of
different types of palustrine wetlands. Intensive afforestation activities in the headwaters of the
Umzimvubu catchment during the past 15 years indicate that the wetland characteristics are changing. To
date, insufficient information is available about the impact of these noticeable land use changes on both the
hydrology of these wetlands and the interaction with their tributary catchments. As a consequence, a
multidisciplinary project has been initiated to identify and assess alterations in wetland dynamics caused
by these land use changes. Consequently, the delineation and inventory of different types of wetlands can
be seen as a prerequisite for the hydrological modelling and the assessment of these phenomena in a basin
Recent studies (LORENTZ & ESPREY 1998, DAHLKE ET AL. 2004) have shown, that the hydrological
dynamics within the Umzimvubu headwaters are mainly controlled by the geological, topographical, and
substrate conditions. As a consequence, different types of wetlands were originated, showing a high
variability due to their spatial distribution as well as their characteristics such as size, topographic position,
duration of saturation, vegetation composition, and morphometric features.
Based on these hydrological and geomorphic characteristics, a hydro-geomorphic classification
approach was developed representing a synthesis of the hydro-geomorphic classification system of
BRINSON (1993) and SEMENIUK & SEMENIUK (1995) as well as a regional approach for South Africa
introduced by LONGMORE (2002). The classification system provides the theoretical framework for a
methodology to delineate wetlands in terms of their hydro-geomorphic characteristics by utilizing a variety
of terrain parameters, which have been calculated from a Digital Elevation Model (DEM). The presented
approach provides an efficient method for systematic inventories of wetlands emphasizing on abiotic
78 Remote Sensing & GIS for Environmental Studies, edited by Stefan Erasmi, Bernd Cyffka, Martin Kappas
(Göttinger Geographische Abhandlungen, Vol. 113, Göttingen, 2005)
The approach has been applied to three subcatchments (Mooi, Wildebees and Gatberg rivers) with a
total size of about 816 km² in the upper parts of the Umzimvubu catchment (see Fig. 1). The geology is
dominated by Triassic sediments belonging to the Karoo Sequence. The manifold geological base results in
a scarpland (Schichtstufenlandschaft) with wide valleys, numerous canyons and series of sloping plateaus.
Regarding the climate the region lies in a summer rainfall area that is characterized by rainfalls from
September to April (MAP: 750 mm) showing a high variability.
Figure 1. Location and characteristics of the study area.
2 WETLAND TYPES AND CLASSIFICATION
The hydro-geomorphic classification approach presented in this study (see Fig. 2) is based on the
hydrologic and geomorphic properties of wetlands. According to the wetland characteristics driven by
relief, a hierarchical, level-based classification framework has been developed. The first level distinguishes
wetlands in terms of their position in the landscape. The second level defines subtypes according their
morphometry, or “geomorphic setting” to consider the flow and storage of water. In the third level these
subtypes are classified regarding their surface hydrological regime, which differentiates surface flow
mechanisms, hydrodynamics, water sources and connectivity to flow channel. Moreover each wetland type
has been related to the duration of the saturation period or hydroperiod based on field survey.
In detail wetlands of the first level are distinguished into the classes plateau, slope and valley bottom
wetlands. These classes integrate differences in topographic position of a wetland in relation to a typical
slope profile as well as its position in the surrounding catchment. In this context the topographic position
of these wetlands is mainly associated to geologic settings of the interbedded sandstone and claystone
layers which have controlling impact on spatial distribution of water.
According to the geomorphic classification of SEMENIUK & SEMENIUK (1995) wetland subtypes are
differentiated of the second level by using attributes of morphometric settings or landforms. Within the
applied classification approach slope and plateau wetlands are distinguished into the two classes “channel”
or “depression” wetlands. These attributes characterize size, shape and depth of wetlands and thus their
potential capability of water retention. Consequently, channel wetlands are often located in areas of depth
contours. Their discharge or lateral inflow from the upslope catchment is channelized within the substrate
layer on or close to the surface. Depression wetlands are more or less basins with variable size and depth,
which are most commonly found in high elevations or in areas that are hydrological isolated from channel
flow. The flat areas of the valley bottom wetlands are divided in “channel” or “floodplain” wetlands,
which mainly differ in size, area of inundation, and flow mechanism like concentrated channel flow or
diffuse substrate flow. Wetland subtypes of the third level are separated in terms of their surface
hydrological regime or types of water sources considering differences in hydroperiod, dominant direction
of water flow, and topographic terrain settings.
Göttinger Geographische Abhandlungen, Vol. 113 79
Figure 2. Hydro-geomorphic classification system of wetlands in the study area.
3 DATA BASE AND METHODS
3.1 DATABASE AND PRE-PROCESSING
Former studies have provided a comprehensive data base, which could be used for the presented study.
Land use maps were available from multi-temporal classifications of Landsat TM/ETM data
(HELMSCHROT & FLÜGEL 2002). Stand characteristics for the forest plantations have been provided by
field studies and from the forest data base. The Environmental Conservation Data Base (ECDB) includes
detailed information due to the land cover/use in the non-forested areas of the North East Cape Forest
(NECF). In addition to that, vegetation and wetland distribution maps, physical soil as well as plant-
physiological parameters and others have been provided by numerous field campaigns.
The main database for the digital terrain analysis is a high quality Digital Elevation Model of ITED-2
(Interferometric Terrain Elevation Data, Level 2) data of the Shuttle Radar Topography Mission (SRTM)
with a resolution of 25 x 25 m². Since a short-wave X-band system has been used within the SRTM, the
data set must be addressed as a Digital Surface Model (DSM) including the height of vegetation and
buildings (KOCH & LOHMANN 2000). Additionally, the radiometrically corrected and calibrated DSM
contains different types of errors resulting from the interferometric technique and the data processing,
which mainly can be found in areas of open water bodies and rough terrain. Therefore, an intensive pre-
processing has been done to eliminate these artefacts and to reduce the vegetation effects.
Since the land use within the study area is dominated by grassland and plantation forestry, the reduction
of the vegetation heights can be limited to the afforested areas. According to the age and the planted
species, the trees exceed heights of 20 m rarely. Based on field measurements and information taken from
the forestry data base, the mean vegetation height for the plantations can be classified as shown in Tab. 1.
A vector layer containing polygons with the different height classes was converted to a raster file in
order to subtract the tree heights from the SRTM DSM in the forest areas. According to spatial
uncertainties and buffer zones in the boundary areas of the plantations, a method for a stepwise adjustment
of the tree heights has been introduced. As schematically shown in Fig. 3, the tree height has been
increased from 0 m up to the stand specific tree height using linear steps from the plantation boundary to
the base zone. The optimum numbers for the height steps as well as the overall distance of the buffer were
iteratively tested by checking the resulting DEM using height profile tracking. Small remaining artefacts
which could not be corrected by calculating height buffers were smoothened by applying a low-pass filter.
As a result the DSM could have been corrected and a Digital Elevation Model (DEM) has been calculated.
The elimination of artefacts concerning the surface roughness and small pits of the DEM (BEHRENS
2000 [ConFit; ESRI/ArcScripts]) has been realized by using a limit-filter with a 3x3 pixel window.
Furthermore, local random errors within the range of open water bodies or steep slopes have been detected
80 Göttinger Geographische Abhandlungen, Vol. 113
by the calculation of the local standard deviation of the DEM. By setting a threshold value to the standard
deviation only those areas with height differences of more than 130 m have been identified. These areas
were revised by interpolating new height values or filling the gaps. Finally, the stream network or flow
accumulation has been delineated from the DEM by applying a phenomenon-based approach of pit
removal (BEHRENS 2002). At first this algorithm locates all sinks in a DEM and burns small segments of a
calculated stream network into the DEM in a predefined circle around the located pits. Consequently, the
refilling of the known depressions in the DEM is reduced to minimum extent by burning-in these segments
with a specified depth. As a result, the DEM shows more accurate results in hydrological applications.
Table 1. Mean heights of Pinus and Eucalyptus stands within the afforestation areas of NECF.
Pinus spp. Eucalyptus spp.
age height age height
plantation period plantation period
[years] [m] [years] [m]
<3 03/1997 - 03/2000 0 <1 03/1999 - 03/2000 0
3–6 03/1994 - 03/1997 5 1–2 03/1998 - 03/1999 3
6 – 10 03/1990 - 03/1994 9 2–4 03/1996 - 03/1998 9
10 – 20 03/1980 - 03/1990 15 >4 before 1996 15
> 20 before 1980 22
Figure 3. Stepwise adjustment of buffer zones in the boundary areas of the plantations to subtract vegetation height
from the SRTM DSM.
3.2 TERRAIN-BASED DELINEATION OF WETLANDS
Based on the developed hydro-geomorphic classification approach an expert system for the terrain-
based delineation of wetlands was developed using different combinations of specific terrain attributes. As
shown in Tab. 2 nine parameters out of 69 terrain attributes (BEHRENS ET AL. 2005) have been used.
Table 2. Selected terrain attributes for the terrain-based delineation of wetlands.
terrain attribute author of calculation algorithm
HORN (1981), DIETRICH & MONTGOMERY
relative profile curvature BEHRENS (2003)
relative planform curvature KLEEFISCH & KÖTHE (1993)
DIETRICH & MONTGOMERY (1998)
JENSON & DOMINGUE (1988)
relative hillslope position BEHRENS (2003)
height above drainage channel BEHRENS (2003)
compound topographic index BEVEN & KIRKBY (1979)
floodplains based on flood
According to the hydrological and geomorphic characteristics of wetlands, the listed terrain attributes
characterize geomorphometric as well as moisture differences within the landscape. Since the
reclassification and combination of these attributes provide a reasonable potential to delineate different
types of wetlands, a hierarchical, rule-based delineation approach has been developed (Fig. 4). As shown,
Göttinger Geographische Abhandlungen, Vol. 113 81
its application is based on the sequential processing of the DEM for the delineation of each wetland type
utilizing their specific terrain features. A detailed description is given in the following sections.
Figure 4. Methodology of terrain based delineation of different wetland types.
3.2.1 Valley bottom wetlands
Valley bottom wetlands are connected to the stream network and are situated in areas of depth contours
within a landscape. Those areas can be delineated calculating a flood simulation of the DEM (BEHRENS
2003) which is based on the terrain attribute “height above drainage channel”. This terrain attribute
calculates terrain height of the surrounding terrain area of a stream network in relation to terrain height of
the stream itself. By setting the height level of the stream to zero all surrounding areas have positive
heights above the stream. The spatial extent of the flooded areas can be limited within the simulation
process by setting values for flood height as well as for the contributing area. That means that the size of
hillslope or catchment area is calculated for each cell of a DEM using the flow accumulation. Best results
for the delineation of all possible valley bottom wetlands within the presented study had been achieved
using a value of 40 ha for the contributing area and a flood height of 1.80 m. These values show a
maximum correlation in comparison to validation wetland areas.
After delineating all potential valley bottom wetlands the resulting raster map was converted to a
polygon-format to classify into “depression” and “channel” wetlands. Therefore the valley bottom
wetlands with a large spatial extent were separated from wetlands with a small linear shape. The
differentiation is based on a buffer around the stream network with a predefined width of 75 m on both
sides of the stream. All wetlands that are lying completely within the predefined buffer are “channel”
wetlands all others were classified as “floodplain wetlands”.
Within the third level of hydro-geomorphic classification the “surface hydrological regime” of all
valley bottom wetlands was classified. According to the classification scheme the hydrological
characteristics of valley bottom wetlands show a close relation to the slope of the terrain. Therefore the
mean slope for each polygon of the valley bottom layer had been calculated using a zonal statistic
algorithm. Now the wetland type “headwater” could have been classified by selecting all polygons with a
first stream order and a mean slope of more than 4 degrees. Wetlands with a mean slope value of 2 – 4
degrees were classified as “middle gradient” wetlands and those with slope values less than 2 degrees as
“low gradient” wetlands.
3.1.2 Slope wetlands
For the terrain based delineation of “slope” wetlands a combination of the four terrain attributes “relative
profile curvature”, “relative planform curvature”, “compound topographic index” and “height above
82 Göttinger Geographische Abhandlungen, Vol. 113
drainage channel” was used. First of all the basins and depression on the slopes were delineated by
reclassifying both curvature attributes. Based on the normal range between -100 for concave areas and 100
for convex area the classification is based on values less than -20 for both attributes. The results were
combined to locate all possible depressions within the slopes. Small sized depressions could have been
classified using the profile curvature while the planform curvature primary localizes linear depressions on
slopes. In a next step the reclassified curvature map was combined with a reclassified map of the
“compound topographic index” (values ≥ 3.25) as moist areas at the footslopes have not been localized
sufficiently by the reclassified curvature map. Additionally the maximum slope was set to 9 degrees (15.8
%), because wetlands never occur on steeper slopes in this area.
To differentiate the morphometric characteristics of the wetlands the raster map with all possible slope
wetlands was converted to a polygon file. This differentiation is based on the shape (round or longish) of
each polygon Based on measuring the superficial area of each wetland polygon a quotient of theoretical
(circle) and true perimeter was calculated. Thus round shapes have a similar value as a circle whereas
longish shapes show higher values. Based on this principle the shape of all slope wetlands was
differentiated into the types “depression” (quotient ≤ 1.75) and “channel” (quotient ≥ 1.75).
For the delineation of surface hydrological regime surface inflow, throughflow, and outflow was
determined for each wetland based on calculating the maximum flow accumulation for each wetland
polygon. Therefore the contributing area, i.e. number of upslope cells draining into a polygon multiplied by
grid size was calculated. Based on field investigations we found that a flow channel starts to develop at a
contributing area of at least 10 ha (160 grid cells). That means that wetlands, whose size of hydrological
catchment is smaller than 10 ha do not have a surface inflow and outflow channel and therefore belong to
wetland type “no surface inlet and outlet”. Wetlands with sizes greater than 10 ha were classified as
wetlands with “surface outlet only” while wetlands with a maximum inflow value of more than 10 ha in
the upslope position and a value of more than 47 ha in the downslope position show a typical throughflow
hydrological regime and are classified as wetland “with surface inlet and outlet”.
3.1.3 Plateau wetlands
According to the hydro-geomorphic classification, plateau wetlands are found on the hilltops and plateaus
respectively within the study area. To delineate these wetlands a combination of the terrain attributes
“height above drainage channel” (sect. 3.2.1) and “relative hillslope position” were calculated. While the
“relative hillslope position” is suitable for dividing stretched slopes the parameter is not suitable in areas
with low relief because of small scale artefacts. In Addition with “height above drainage channel” all
plateau areas were delineated using values of more than 15 m for “height above drainage channel” and
more than 1,2 for “relative hillslope position” (0.0 is mid-slope). All potential plateau wetlands were
localized by delineation of depressions based on calculation and reclassification of the terrain attribute
“relative profile curvature”. As described in sect. 3.2.2 the parameter is very sensitive for local changes in
terrain .Thus small depressions with a size of minimal 4 pixels (2500 m²) could be delineated, based an a
threshold of -10.
Within the seconded classification level each wetland polygon was reclassified into the two wetland
types “depression” and “no depression” based on the “depth” of the depression. Wetlands localized in deep
concave depressions (curvature values ≤ -15) are assigned to the class “depression” while wetlands
positioned in concave to intermediary shaped areas (curvature vales -10 – -15) are classified as “no
The differentiation of “surface hydrological regime” was performed in the same way as for slope
wetlands as described in sect. 3.2.2. All plateau wetlands are too small in size (< 32 pixels) were classified
as “no surface channel inflow or outflow” which confirms the field observations within the study area.
3.3 ACCURACY ASSESSMENT
The delineated wetland areas have been validated using field survey, detailed land use information given
by the Environmental Conservation Database (ECDB) as well as Landsat TM/ETM data.
The accuracy of terrain based delineated wetlands has been compared with the areal distribution of
wetland areas, which could be extracted from the ECDB. Consequently, both data sets were converted into
raster files in order to apply a pixel-based accuracy determination algorithm based on a confusion matrix
(BEHRENS et al. 2005, in press). Accuracy (AC) for delineated wetlands has been calculated based on a
confusion matrix (Table 3) and the given equation (1).
Göttinger Geographische Abhandlungen, Vol. 113 83
Table 3: Confusion matrix used for validation.
field maps by ECDB rr + ff rr + ff
AC = = (1)
true false rr + ff + fr + rf N
true rr rf
false fr ff with N as total account of pixels of
As a result, the confusion matrix shows an overall accuracy (AC) of 94 % for wetland types of level 1.
The accuracy assessment indicates that the total wetland area could be delineated by the analysis approach.
Since the accuracy assessment provides insufficient information on the spatial distribution of each wetland
type, a field-based validation has been carried out. Therefore, maps with test areas for all wetland types
have been produced. These wetland maps have been validated by field mapping utilizing GPS survey as
well as vegetation and soil analysis. The estimated accuracy of about 90 % refers to a good consistency of
the delineated wetland with real world conditions. Additionally a visual comparison of delineated wetlands
with Landsat TM data acquired on 9 of April 1999 has been carried out. Since remote sensing data reflect
soil moisture and vegetation conditions on a specific date, those data can only be used as supplement
information for accuracy assessment. Nevertheless, the comparison due to the spatial distribution based on
an overlay of the wetland map confirms the achieved accuracies especially for the large valley bottom
4 RESULTS AND DISCUSSION
As a result a wetland distribution map has been provided to be used for further application in hydrological
studies. As shown in Tab. 4 the delineated wetlands cover a total area of about 124 km² in size which
represents 15 % of the study area. Based on the differentiation of the wetland types a detailed inventory of
distributed wetland types was investigated using information about their areal extent. Valley bottoms
wetlands cover an area of 71 km² (57 % of whole wetland area), while slope wetlands cover an area of
about 47 km² (37 %). Minor parts are covered by plateau wetlands, which share only 5 % of the whole
wetland area, i.e. 7 km² in size. The analysis of the level 2 and 3 wetland types indicate, that valley bottom
wetlands with “low gradient landform” show the largest distribution with 34 % (≈ 43 km²), while larger
plateau and slope wetlands are mainly characterized by channelized outflows. The smallest areas of 3 % (≈
3 km²) are represented by slope and plateau wetlands, which are developed in depressions and show no
indication of surface inlet and outlet.
According to the accuracy assessment described in Sec. 3.3, some inconsistencies could be found within
the slope wetland class. Such differences between delineated and mapped wetlands mainly occur within the
boundary area of plantations. This can be explained by the increasing influence of grid resolution on the
delineation approach. On the other hand, inconsistencies could be found due to an additional, non-
referenced wetland type. This wetland type has been originated by emergent inflow from sandstone or
claystone layers. Basically these hydrological isolated, small-patched wetlands do not indicate a
topographic depression and therefore could not be delineated by terrain attributes itself. A delineation of
these geologically controlled wetlands might be possible utilizing additional geological data and need to be
tested in further studies.
The most problematic wetland types in terms of accuracy are plateau wetlands which are highly
variable due to their small size and short term saturation. Since those wetlands are situated in small or flat
depressions, they can be rarely delineated by the analysis of SRTM DEM. Consequently, the minimum
size of a plateau wetland should be limited to 2,500 m² (4 pixels) in terrain-based analysis. Although field
survey indicates smaller plateau wetlands, those patches have minor influence on the mesoscale wetland
inventory according to their spatial and temporal dynamics.
Furthermore the validation and field survey have shown, that forest plantations might interfere the
terrain based delineated wetland area especially on slopes and foothills. In those areas the potential land
use should be defined as wetland area, while the actual land use is characterized as plantation forestry. In
most cases it could be found, that this phenomena can be associated to the planting of trees close to or
within temporary wetland areas. Consequently, all wetland areas overlapped by forests have been
identified and subtracted from the original wetlands layer for the later hydrological modelling using GIS
analysis tools. As a result a corrected dataset of wetland distribution could be provided (see Tab. 4).
84 Göttinger Geographische Abhandlungen, Vol. 113
Table 4: Size of delineated wetland areas in comparison to size of wetlands excluding planted areas.
wetland areas wetland areas excl. planted areas
swa pwa psa swa pwa psa
plateau 6.6 km² 5.3 % 0.8 % 4.9 km² 4.7 % 0.6 %
slope 46.5 km² 37.5 % 4.6 % 35.4 km² 33.6 % 4.3 %
valley bottom 71.1 km² 57.2 % 8.7 % 65 km² 61.7 % 8.0 %
total 124.2 km² 100 % 15.2 % 105.3 km² 100 % 12.9 %
swa …size of wetland area; pwa …percentage of wetland area; psa …percentage of study area
Since the provided wetland distribution map represents the potential wetland area, it needs to be taken
into account, that wetlands are highly dynamic elements of the landscape. Their spatial extent and
occurrence (saturation, water table, vegetation, etc.) are significant related to climatic conditions, which
can verify over years or months (see Fig. 5). The flexibility of the method allows the consideration of
respective changes by an adjustment of parameters, thresholds and combinations within the delineation
process. Nevertheless, the applied approach provides an excellent data base, which will be used for further
Figure 5: Spatial distribution of terrain based delineated wetlands and wetlands extracted from the ECDB.
The presented study introduces a methodology to delineate and inventory palustrine wetlands using a
terrain-based analysis approach. Based on intensive field studies within 3 headwater catchments of the
semi-arid Eastern Cape, South Africa, a hierarchical hydro-geomorphic classification was developed which
provides the theoretical framework for the terrain based delineation of hydrologic and geomorphic
characteristics of wetlands. In general the hydro-geomorphic classification system differentiates plateau,
slope and valley bottom wetlands which differ in terms of surface hydrological characteristics, terrain
position as well as morphometric features. This framework has been used to develop an expert system for
the operationalized delineation of wetlands by a set of different combinations of terrain attributes
calculated from a high quality SRTM Digital Surface Model (DSM) of 25x25 m² grid size.
First of all the DSM has been corrected in terms of systematically errors and effects by vegetation
heights utilizing detailed spatial information about plantations and tree heights. Furthermore a variety of
terrain attributes such as several curvatures, the compound topographic index as well as hillslope position
parameters has been extracted from the corrected Digital Elevation Model (DEM). It could be shown, that
a terrain-controlled flood simulation according to the approach of BEHRENS (2003) showed best results for
delineation of valley bottom wetlands. Slope and plateau wetlands were delineated using the profile and
planform curvature or a combination of them and the compound topographic index. Terrain position of
slope and plateau wetlands has been differentiated using the terrain attributes “height above drainage
channel” and “relative slope position” that describe height and distance of terrain to stream channel.
Morphometric differentiation of wetland types is based on shape characteristics, while differentiation of
the surface hydrological regime was distinguished using flow accumulation. As a result of the rule-based
delineation approach and the accuracy assessment, a wetland inventory has been done representing the
distribution and extent of the specified wetland types in the study area. It could be shown, that the wetland
Göttinger Geographische Abhandlungen, Vol. 113 85
areas are dominated by larger valley bottom wetlands and channelized slope wetlands. As shown,
uncertainties can be associated to the quality of the DEM and its resolution as well as interfering land use.
The study has shown the potential of the presented approach for wetland inventory utilizing terrain
parameters developed from a SRTM-based DEM. Due to its flexibility the approach can be used for
numerous applications related to the analysis of terrain features for environmental assessment and
modelling. Since the procedure as such can be adjusted by utilizing further terrain attributes such as terrain
based stream-power and sediment transport capacity indices (MOORE ET AL. 1991, MONTGOMERY &
DIETRICH 1992), indices for landscape assessments of erosion (FELDWISCH 1995) or compound indices to
estimate spatial and temporal distribution of solar radiation (DUBAYAH & RICH 1995, BEHRENS 2003) the
method might be improved to specific research needs.
This study is part of the project “Integrated landscape model of wetlands and its impact for the water cycle
of semi arid river basins at example of the Umzimvubu catchment, Eastern Cape, South Africa” funded by
the German Research Foundation (DFG). Acknowledgements are also given to Dr. Simon Lorentz
(University of Natal) as well as Mondi Forests Ltd. for data and field support.
1 Lorentz, S.A., Esprey, L.J., 1998: Baseline hillslope study prior to afforestation in the Umzimvubu headwaters of
the North East Cape Province, South Africa. In: Kovar, K., Tappeiner, U., Peters, N.E., Craig, R.G. (eds.):
Hydrology, Water Resources and Ecology in Headwaters (Proceedings of the HeadWater '98 Conference held at
Meran/Merano, Italy, April 1998, IAHS Publ. no.248, pp. 267- 273.
2 Dahlke, H., Helmschrot, J., Flügel, W.-A., 2004: Ein integrativer Ansatz zur landschaftsbezogenen Typisierung
von Feuchtgebieten im Einzugsgebiet des Umzimvubu, Ostkap-Provinz, Südafrika. Zbl. Geol. Paläont. Teil I,
Heft1/2, in press.
3 Brinson, M.M., 1993: A hydrogeomorphic classification for wetlands. U.S. Army Corps of Engineers,
Washington, D.C., WRP-DE-4.
4 Semeniuk, C.A., Semeniuk, V., 1995: A Geomorphic Approach to Global Classification for Inland Wetlands.
Vegetatio 118: pp. 103-124.
5 Longmore, J., 2002: The geomorphology of wetland systems, Upper Mooi River catchment, South Africa. MSc.-
Thesis at the University of Natal, Pietermaritzburg, South Africa.
6 Helmschrot, J., Flügel, W.-A., 2002: Land use characterisation and change detection analysis for hydrological
model parameterisation of large scale afforested areas using remote sensing. - Physics and Chemistry of the Earth
27, pp. 711-718.
7 Koch, A., Lohmann, P., 2000: Quality assessment and validation of digital surface models derived from the
Shuttle Radar Topography Mission (SRTM). Proc. ISPRS, Vol. XXXIII, Amsterdam.
9 Behrens, T., 2000: ConFiT – A Convolution Filter Tool for ESRI-Grids. Internet Referenz:
10 Behrens, T., Scholten, T., 2002: DGMK 20+. Erstellung der Digitalen Geomogphographischen Karte für das
Bundesland Rheinland-Pfalz im Maßstabsbereich 1:25.000 bis 1:50.000. Dokumentation. Geologisches
Landesamt Rheinland-Pfalz. Mainz.
11 Behrens, T., Förster, H., Scholten, T., Steinrücken, U., Spies, E.-D., Goldschmitt, M., 2005: Digital Soil Mapping
using Artificial Neural Networks. Journal of Plant Nutrition and Soil Science, in print.
12 Horn, B.K.P., 1981: Hillshading and the reflectance map. Proc. IEEE 69(1): pp. 14-47.
13 Dietrich, W.E., Montgomery, D.R., 1998: A digital terrain model for maping shallow landslide potential. Internet
14 Behrens, T., 2003: Digitale Reliefanalyse als Basis von Boden-Landschaftsmodellen: Am Beispiel der
Modellierung periglaziärer Lagen im Ostharz. Diss., Justus-Liebig-Universität, Inst. für Geographie, Giessen.
15 Kleefisch, B., Köthe, R., 1993: Wege zur rechnergestützten bodenkundlichen Interpretation digitaler Reliefdaten.
Geologisches Jahrbuch F 27: pp. 59-122.
16 Jenson, S.K., Domingue, J.O., 1988: Extraction topographic structure from digital elevation data for geographic
information system analysis. PE&RS 54(11): pp. 1593-1600.
17 Beven, K.J., Kirkby, M.J., 1979: A physically-based, variable contributing area model of basin hydrology.
Hydrological Sciences Bulletin 24: pp. 43-69.
18 Moore, I.D., Grayson, R.B., Ladson, A.R., 1991: Digital terrain modelling - a review of hydrological,
geomorphological and biological application. Hydrological Processes 5: pp. 5-30.
19 Montgomery, D.R., Dietrich, W.E., 1992: Channel initiation and the problem of landscape scale. Science 255: pp.
20 Feldwisch, N., 1995: Hangneigung und Bodenerosion. Boden und Landschaft 3.
21 Dubayah, R., Rich, P.M., 1995: Topographic solar radiation models for GIS. Int. J. Geographical Information
Systems 9: pp. 405-419.
86 Göttinger Geographische Abhandlungen, Vol. 113