694 
Figure 2. Use of Landsat imagery in catch 
ment modelling. 
In most models which use Landsat data the spatial 
variability of the hydrological processes is ignored. 
However, in catchments containing complex patterns 
of recharge and discharge areas, Landsat imagery can 
be of great use in quantifying the field heteroge 
neity in a semi-distributed model. This is especial 
ly true for inaccessible mountainous areas. 
2. METHODOLOGY 
A phased methodology is suggested for the régionali 
sation and runoff prediction in heterogeneous Alpine 
areas (Figure 3). 
I Reconnaissance stage: Identification of broad physio 
graphic zones (maps and Landsat MSS imagery 1:200.000) 
II Mapping, field surveys and modelling in a reference 
area 
1. Identification of patterns of land units 
- visual interpretation of Landsat MSS and TM 
1:100.000 - 1:25.000) 
- digital classification of Landsat MSS and TM 
2. Hydrological quantification of land units, resul 
ting in hydrological land units 
- field observations 
3. Synthesis of hydrological land units in flou modal 
- systems analysis 
III Simulation and verification of the model in a control 
area 
Figure 3. Methodology. 
2.1 Reconnaissance stage : Identification of broad 
physiographic zones 
In the reconnaissance stage the study area is re 
stricted to one broad physiographic zone of a few 
thousand to tens of thousands of square kilometres. 
No hydrological information transfer is allowed 
from one zone to another. 
Within a climatic zone these physiographic zones 
are mainly distinguished on basis of their lithology 
and structural geology. The general morphology is 
used for further differentiation (Meyerink, 1976). 
Apart from information from small scale maps 
(1:200.000), the visual interpretation of Landsat 
MSS imagery is an important data source. 
2.2 Mapping, field surveys and modelling in a refer 
ence area 
In the next stage a reference area is selected which 
is expected to contain all the hydrological diversity 
within that particular physiographic zone. A hier 
archical mapping and modelling procedure is per 
formed in this reference area (Figure 3). 
On the lowest level patterns of landsurface- 
physical features (land units) are identified. For 
the delineation of vegetation and landuse units a 
digital classification of Landsat MSS and TM data is 
pursued. The identification of geomorphological fea 
tures, fault patterns, etc., is performed on basis 
of the visual interpretation of Landsat MSS and TM 
imagery (1:100.000-1:25.000). 
Land units derived from imagery are checked in the 
field and quantified hydrologically, based on hydrol 
ogical characteristics like the generation of peak 
discharge or baseflow, sediment yield, soil moisture 
content, etc. Thus, the land units are converted to 
hydrological land units. 
These two-dimensional hydrological units are com 
posed by a systems analysis into interconnected flow 
systems of surface waters, soil waters and ground- 
waters. Groundwater flow systems are e.g. subdivided 
in deep regional, deep to intermediate subregional 
and shallow local systems (Engelen, 1984). 
For the identification of regional groundwater 
flow systems, which cross main surface water divides, 
Landsat MSS imagery is of great use (scale 
1:100.000). On this system level regional fault pat 
terns are analysed. Subregional and local flow sys 
tems, within the major subcatchments, are mainly de 
lineated on basis of geomorphological features and 
vegetation and land use patterns. Landsat MSS and TM 
images (scale 1:50.000-1:25.000) are an important 
tool for identifying these features. 
Generally, for high Alpine environments with 
fractured or karstic rocks the recharge takes place 
over broad areas, while the outflow of the systems 
is limited to concentrated spring areas. The rela 
tively poor resolution of Landsat MSS and TM is a 
severe limitation for the identification of concen 
trated spring areas. Furthermore, spring areas are 
mostly not uniquely related to a specific vegetation 
type. Thus one has to rely on field observations and 
aerial photographs (scale 1:25.000-1:10.000). 
The hydrological character of a catchment is not 
only modelled by the conventional lumped parameters 
and variables, but in this semi-distributed flow 
model, the spatial distribution of the hydrological 
land units, like distance to outlet, area, mean 
height and aspect, is also quantified. This spatial 
information, derived from a combination of Landsat 
imagery maps and field surveys, is stored in a 
Geographic Information System (G.I.S.). Although the 
desired grid scale of this data bank depends on the 
variable and the type of region to be modelled, for 
most input data a grid scale of 30 m (resolution 
TM), or even 80 m (resolution MSS), seems sufficient 
(Hendriks, in prep. : study in East-Luxembourg). 
The response time and the initial state of water 
contents of the various flow systems are obtained 
by hydrograph analysis, analysis of spring dis 
charges and soil moisture measurements. 
Some model parameters of the hydrological land 
units, like the interception storage capacity, have 
to be measured in small field plots. The field-data- 
based parameters may be correlated with reflectance 
indices of Landsat MSS and TM. 
2.3 Simultation and verification of the model in a 
control area 
In the last stage of the investigation for a control 
area the patterns of land units are delineated from 
Landsat imagery. On basis of the gained hydrological 
quantification of these land units a prediction is 
made of the types and the spatial distribution of 
hydrological land units and flow systems. The stream 
runoff and (potential) sediment yield of the control 
area is predicted by the developed semi-distributed 
flow model. The mapping and modelling results are 
verified by fieldwork and aerial photography 
(1:25.000). 
3. CASE STUDY 
The three-phased methodology is presently being ap 
plied to the complex Alpine environment of the N- 
Italian Dolomites (Figure 4). The study forms part 
of an extensive research and graduate training pro 
gram on mountain hydrology in the Dolomites, carried 
out since 1966 by the Department of Hydrogeology and 
Geographical Hydrology, Institute of Earth Sciences> 
Amsterdam F: 
1974; Seyhai 
The broad 
Liassic roc) 
selected as 
and the spr: 
and discharc 
the Upper-B< 
trol areas ; 
Legend : 
Figure 
3.1 Reconnai 
The heteroge 
Liassic rock 
three main f 
1. Litholc 
present in t 
sandstones a 
masses are f 
of marls vol 
ential erosi 
out 1000 a 1 
ons. 
2. Structu 
stone) forma 
tectonics, w 
(Engelen, 19 
ween the ree 
shales and g 
pirically by 
This resulte 
rock types. 
3. Quatern 
tures, glaci 
talus cones 
mass-movemen 
rocks gave t 
Most geologi 
identified w 
The mean a 
1600 m is ap