Basic HTML Version
Understanding the spatial and temporal occurrence of landslides using satellite and airborne technologies: Papua New Guinea
5
Fig. 5|
Landslide susceptibility maps for the Western Province case study area using models; (A) with control factors, including slope,
lineament distance, lithology and elevation, (B) with control factors, including slope, lineament distance and lithology, (C) with controls
factors, including elevation, aspect and drainage distance, and (D) with control factors, including slope, lineament distance, lithology,
drainage distance and elevation. Observed landslides are shown by black, dashed polygons.
seated landslides, whose morphological characteristics are easier to identify using high resolution digital elevation
models (DEMs; Van Den Eeckhaut et al., 2005; Lin et al., 2004), could be identified. The use of a range of remote
sensing approaches dramatically increased the sample size of landslides available for susceptibility analysis (n=191
landslides in the Western Province case study area; n=366 landslides in the Chimbu Province case study area).
These two remote sensing datasets also made it possible to distinguish landslide head scarps from depositional
features, which is particularly important for accurate susceptibility analysis.
The location of the identified landslide head scarps in each case study region were assessed against the spatial
distribution of lithological (lineaments, drainage, rock type and land cover) and topographical (elevation, slope,
aspect and curvature) characteristics. The control factors of greatest relevance to landslide susceptibility in each
case study area were identified based on a ratio between topographic and lithological parameters at landslide
head scarps and topographic and lithological parameters at non-landslide sites (Coe et al., 2004). It was found
that of the topographic parameters, slope was the most important factor for landslides in Western Province, while
elevation played a more significant role for landslides in Chimbu Province. In terms of the lithological parameters,
the distance from lineaments was found to be very instrumental for landsliding in both case study areas, as was
the lithology, with high ratio values being calculated for mudstone in Western Province and limestone in Chimbu
Province.
In order to generate a single landslide susceptibility map for each case study area, those parameters considered
most important for landsliding in each area needed to be combined, so that areas of high landslide susceptibility
could be distinguished fromareas of low/no landslide susceptibility. Fuzzy logic was used for this purpose, combining
various control factor datasets to produce different landslide susceptibility maps (Pradhan, 2011). Four separate
susceptibility models and maps were tested for each case study area (Fig. 5) and examined based on their ability
to accurately ‘forecast’ the locations susceptible to landslide initiation, while not producing too many false alarms.
The most favourable susceptibility model for Western Province used slope, lineament distance and lithology as its
control factors (Fig. 5(B)), while the most favourable model for Chimbu Province used elevation, lineament distance
and lithology as its control factors.