Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
A methodology for integrating satellite imagery and field 
observations for hydrological régionalisation in Alpine catchments 
R.Allewijn 
Department of Hydrogeology and Geographical Hydrology, Institute of Earth Sciences, Free University, Amsterdam, 
Netherlands 
ABSTRACT : Years of intensive fieldwork in the N-Italian Dolomites have demonstrated that in complex Alpine 
environments predictions of stream runoff and sediment yield for ungauged watersheds require a semi-distrib 
uted, physically-based régionalisation model. The physically-based character of the model has to be guaranteed 
by detailed field observations, while Landsat remote sensing data can be quite valuable in quantifying the 
distributed nature of the model. 
A hierarchical régionalisation procedure is presently being applied to the N-Italian Dolomits : 
(1) A broad physiographic zone of Permo-Triassic-Liassic rocks has been delineated with Landsat MSS images 
(1:200.000). 
(2) Vegetation and landuse units are identified by a supervised digital classification of Landsat data. 
Other patterns of landsurface-physical features are delineated by the visual interpretation of Landsat MSS and 
TM imagery (1:100.000-1:25.000). 
(3) After an insight and quantification is gained of the relation between the landsurface-physical variables 
and the hydrological character of the Permo-Triassic-Liassic rocks, by comparison with hydrological field sur 
vey data, surface water systems and related groundwater flow systems are identified. 
(4) The spatial characteristics of the hydrological units are stored in a Geographic Information System, 
which sërves as a data bank for semi-distributed water and sediment yield models. 
(5) Landsat data can further be used to correlate reflectance indices with specific field-data-based model 
parameters. 
This procedure is being developed for a reference area and will be tested for a control area during a later 
stage of the investigation. 
In conclusion, if one knows how to use remotely-sensed information, this data source could be an important 
additional tool in solving the hydrological régionalisation problem. 
1 INTRODUCTION curve Number Model (Soil Conservation 
Service, 1972) cannot be applied to catchments 
1.1 Problem-oriented régionalisation approach 
with a complex spatial distribution of hydrologi 
cal units. A more realistic way to deal with the 
In Alpine environments, where a detailed gauging 
network is often missing, reliable predictions of 
stream runoff and sediment yield are most needed. 
The extrapolation of records to ungauged catchments 
can be performed by several régionalisation methods 
(Figure 1). The choice for a specific approach de 
pends on the nature of problems to be solved, the 
scale at which a solution is required and the com 
plexity of the research area (Simmers, 1984). 
meters have no unique relation to field measure- 
ments of hydrological phenomena or to landsurface- 
physical variables, are not suited for the extrapo 
lation of the results of gauged watersheds to un 
problem of heterogeneity is to divide the catchment 
into a number of homogeneous subareas and to model 
the hydrological processes for each subarea separa 
tely (semi-distributed models such as the USDAHL-74 
Model;- Holtan et al., 1975). 
Calibrated catchment models, in which the para- 
regionalisation models 
gauged watersheds. Consequently, in complex Alpine 
regions, a physically-based model with a (semi)dis 
tributed character is recommended. 
[statistical models| [conceptual modelsj 
1.2 Use of satellite imagery in catchment modelling 
Figure 1. Régionalisation models. 
Until now, the evaluation of the applicability of 
remote sensing techniques to infer basic model para 
meters has been focused on existing catchment models 
(Peck et al., 1981; Engman, 1982; Rango, 1985). How 
ever, these models do not have a significant poten 
tial for using remotely-sensed information (Peck et 
al, 1982). 
Lumped models, such as the Stanford Watershed 
Model IV (Crawford and Linsley, 1966) and the SCS 
In (Alpine) regions, where a large number of factors 
controls the hydrological regime, it seems to be 
very difficult to predict the runoff of ungauged 
catchments by purely statistical methods (Mosley, 
1981; Ebisemiju, 1979). In such areas a conceptual 
model is more appropriate, in which the types of 
hydrological processes and their interrelationship 
with landsurface-physical variables are considered. 
The incorporation of Landsat data in catchment 
response modelling has been described, among others, 
by Ragan and Jackson (1980 : SCS Curve Number Model), 
Groves and Ragan (1981 : Modified Stanford Watershed 
Model) and Fisher and Ormsby (1982 : USDAHL-74 
Model). As the Landsat bands can give only indirect 
estimates of hydrological parameters, one has to 
determine how landsurface-physical features are re 
lated to hydrological characteristics of the area. 
Field observations are indispensable for this pur 
pose (Figure 2). 
693