facilitates the interpretation compared to 2D on screen
perception. However, the task should be entrusted to an
experienced geologist and followed by a ground validation
through a field survey. Another point to stress out is that the
geohazard inventory should be regularly updated using multi-
temporal images and preferentially after each major rainfall or
earthquake (Wunderle, 2002). In tropical zones, like Matagalpa,
this action can be seriously hindered because of cloud cover.
Benefits of the SPOT-5 mapping tool compared to traditional
field or aerial methods include the immediate compatibility
with a GIS and the relative quickness of obtaining a global
landslide inventory for large areas without accessing all of the
area of interest.
Semi-automatic approaches, based on image radiometry remain
mainly empirical and their relevance should be assessed from
case to case. At Matagalpa, these methods seem not to be
appropriate for landslides, because they were recognized on
SPOT-5 images more by their morphological features than their
reflectances and their deposits are only weakly contrasted. In
contrast, debris flow events can presumably be detected using
vegetation indexes such as NDVI (Liu et al, 2002) At
Matagalpa, their deposits, as poorly vegetated areas have
slightly positive NDVI values, comprised between 0.0 and 0.2,
similar to the ones of settlements or road infrastructure. The
application of a filter, such as a buffer zone along the drainage
network allows to reduce this confusion and to sort a low
number of candidate areas where debris flows could have been
deposited. Such an easy-to-implement method, not functional
on its own is supportive in validating visual interpretation.
3.2 Susceptibility maps
The input of a pseudo-color SPOT-5 image and derived land
cover classifications to hazard susceptibility mapping appears
as restricted, since the prime factors controlling the
predisposition of soils to landslides, mud and debris flows, such
as geology, terrain slope and hydrography are themes hardly
retrievable. Knowing that hill slopes are the most important
data for susceptibility issues, a DEM from which a classified
slope map can be calculated is a pre-requisite product for this
procedure. This slope map should be applied to filter out the
output of the thematic layers, so that only areas with moderate
slopes (e.g., between 8.5 and 60 degrees, Fig. 3) are kept as soil
candidates with predisposition to natural hazards.
In tropical regions, like Matagalpa, the bedrock geology and the
quaternary units are mostly masked by the vegetation cover, but
the presence of lineaments (surface expression of a basement
block boundary or fault) can be easily detected on a pseudo-
color SPOT-5 image. The lineaments we traced show quite a
similar pattern compared to the published data (faults) of
Havlícek et al. (2002). Their detection is very helpful for a
susceptibility analysis, since high density lineament areas are
considered as prone to landslides, especially when earthquake-
induced. In contrast, SPOT-5 imagery despite its resolution is
not adapted for delineation of secondary and tertiary
hydrographic networks, features that are critical in our
methodology.
Attempts were made to extract from land cover classifications
thematic information corresponding to second-order risk
factors. Annual crop lands, grounds that have properties close to
barren soils are potential soil candidates with a predisposition to
mud and debris flows. The superimposition of 1998 occurred
mud and debris flow events (Cannon et al., 2001) to a slope-
filtered SPOT-5 land cover map (Fig. 3) is in broad agreement
with this assumption, with 25 percent of the Mitch-related
phenomena having actually started from this soil type. Other
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
276
mud/debris-flow-favorable land cover classes seem to be
agricultural and pasture lands at the external boundaries of
forests (Fig. 3) - areas apparently recently deforested. This
suggests that land use changes are also a SPOT-S-retrievable
parameter to take into consideration for susceptibility mapping,
especially when multi-temporal images are available.
3.3 Vulnerability maps
High-resolution SPOT-5 products could be a relevant source of
information in the preparation of vulnerability maps, and in
particular for spatial data collection of landslide or mud/debris
flow-threatened elements. With semi-automatic methods, like
Matagalpä 0 0.4 0,8
(horthern suburbs) itométes
d,
Forest ME Burnt land
Natural vegetation (bush) Eg Urban (settlements, roads)
F ; . e
“| Agroforest (coffee trees)
| Agricultural land . | Filter masking areas with
: 3 ; os «$8 $95 9?
Bare soil (agricultural land) slopes «8.5" and 760
Annual crop = | Debris and mud flow
initiation points
Pasture : I
Figure 3. Overlay of the initiation points of 1998-active mud
and debris flows (Cannon et al., 2001) on a slope-
filtered SPOT-5 land cover map, showing that, north
of the city of Matagalpa, annual crop lands and
deforested areas (pastures and agricultural lands
close to the forest boundaries) are soils with
moderate to high predisposition to geohazards.
Inter]
———
g.
&
Table 1. Summary table with benefits and limitations of SPOT-5 derived products for hazard inventory mappin