stanbul 2004 
ima RS from 
ivers, lakes, 
aps and the 
ers (lateritic 
ased on the 
‘un out areas 
nnon et al, 
th of the city 
andard map 
analysis of 
hazards that 
ion and city. 
Carrefio and 
Lal, 2002). 
id Schneider 
shearing of 
ig a failure 
al mass of 
lar to a fluid 
nix of solid 
jiscous fluid 
r the gravity 
'as mud and 
y driven by 
es are slow 
few cm/y if 
ating phases 
1 subsequent 
igalpa). Key 
| to terrain 
ice of water 
harge zones) 
factors that 
barren soils, 
ards: (a) by 
ventory map 
s of terrain 
ding to their 
usceptibility 
| Sectors are 
'ognition on 
| of terrain 
ed drainage, 
The medium 
e, processed 
EM for 3D 
ni-automatic 
ly the use of 
on-vegetated 
rther filtered 
Sits. 
tegration of 
leveloped by 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
EPFL/GEOLEP in the frame of the CADANAV project 
(CArtographie des DAngers Naturels du canton de Vaud; 
Schneider, 2001). This methodology consists of four steps: a) 
information (DEM, geological and topographic maps, aerial 
photographs, remote sensing data) management aimed at the set 
up of classification maps based on risk factors (geology, slopes, 
hydrographic network distance, water saturation of soils); b) 
overlap of step 1-created theme maps and spatial analysis of the 
predisposition, organized in classes, of soils (geotypes) to 
documented natural hazards; c) validation of predispositions 
calculated under the former step by comparison with field- or 
remote-documented geological hazards; and d) validation of 
potential risk areas (susceptibility map) by comparison with 
ground observation of active risk areas. In this perspective, we 
consider the ability to create from SPOT-5-processed products 
thematic maps (hydrography, barren soils, deforested areas, 
informal settlements) that can be further class-organized 
according to the predisposition of their elements to landslides, 
mud and debris flows. 
The final step to undertake in a full risk assessment is to 
confront the former inventory and susceptibility hazard layers 
with vulnerability maps to calculate the risk. Vulnerability can 
be defined as the probable cost of damage the hazard has done 
or might do to various types of installations. Thus, we look at 
what threatened elements can be retrieved from SPOT-5 
products and how current high-resolution satellite imagery can 
be beneficial. 
3. RESULTS 
3.1 Inventory maps 
On a pseudo-color SPOT-5 image draped over a DEM for 3D 
simulation, several features typical of landslides that have 
actually taken place were visually recognized: a) spreadings at 
toe, with river diversion; b) slope ruptures created by scarps; c) 
   
    
     
  
  
  
   
km-long main 
landslide scarp 
deformation of linear features, like roads (Fig. 2). The 2.5 m 
pixel resolution appears as a significant improvement of SPOT- 
4 imagery, allowing a more reliable detection and the tracing of 
the boundaries of the landslides. Our remote sensing-based map 
shows a good correlation with the field observations collected 
by Havlícek et al. (2002). The main difference lies in the fact 
that large-scale sliding structures, such as kilometer-wide slope 
instabilities are better evaluated using imagery. Some of the 
detected landslides are presumably still active, as revealed by 
low-vegetated parcels or barren soils. This suggests that the 
SPOT-S product can be used for qualitatively monitoring the 
landslide activity, although it is unlikely to determine the nature 
of the movements occurring, i.e. if they are rotational, 
translational or multiple. Limited time gap and atmospheric 
perturbations in our SAR interferometric dataset prevented us 
from corroborating these land deformations. 
The inventory work becomes more delicate when dealing with 
mud and debris flows. Traces of 1998-debris flow deposits have 
almost completely disappeared, except in lower-gradient 
valleys such as Waswali, 5 km west of Matagalpa. The later 
deposits are characterized by large amounts of materials 
accumulated, channel shapes, and the lack of vegetation. 
Similarly, mud flow events, launched by Mitch rainfalls and 
identified on 1998 aerial photographs by Cannon et al. (2001) 
have not been recognized on the 2003 SPOT-5 image, probably 
because new vegetation has already covered the devastated 
areas. Alternatively, mud flow deposits are weakly contrasted 
so that they can be confused with other geomorphological 
features. This underlines that, in the Matagalpa region, only an 
image taken shortly after the damage is suitable for monitoring 
these phenomena. 
From a technical viewpoint, the optimal size for recognition of 
terrain instability is about 1 sq km. The 3D simulation, with the 
draping of the pseudo-color image over a DEM, a process that 
is comparable to a stereographic photo-analysis, greatly 
east-diverted 
stream 
vertical exaggeration is 1.9 
Field data (Havlicek et al. 2002) 
“==... polygenetic landslides 
um. landslide scarps 
—P block creep 
Figure 2. Pseudo-color SPOT-5 image draped over DEM for 3D simulation with the main scarps of large landslides. Typical 
indicators for landslides are disturbed vegetation and deflection of stream traces at the foot of the zone of deposits. Field data of 
Havlièek et al. (2002) are reported for comparison.