Basic HTML Version
Understanding the spatial and temporal occurrence of landslides using satellite and airborne technologies: Papua New Guinea
4
frequencies were then applied to the Bayesian methodology (Berti et al., 2012) and used to calculate the landslide
probability, based on changing magnitudes and durations of rainfall events.
The successful production of these landslide probabilities means that as rainfall forecasts and antecedent conditions
(rainfall which has already fallen) spatially and temporally change, forecasters would be able to identify different
locations, within PNG, which may become more or less prone to landsliding. The application of these thresholds
needs to be carefully considered by forecasters, geotechnical experts, additional stakeholders and end users to
ensure that the right actions are initiated at the right probability threshold.
Landslide susceptibility in PNG
To fully understand landslide hazard, consideration of both the potential triggers (as outlined above) and the
environmental control factors is essential. Therefore, in addition to the development of the probabilistic rainfall
thresholds, a baseline understanding of the landslide susceptibility across PNG was required (i.e. landslide
susceptibility without any input from a rainfall trigger). Landslide susceptibility is broadly defined as the likelihood
of a landslide occurring in a specific area based on the local terrain conditions of that area (Brabb, 1984) and,
therefore, requires a detailed analysis of the lithological and topographical controls of landslides in the area of
study. This is typically completed as a two-step process whereby, (1) landslides are identified and classified within a
historical landslide inventory and (2) environmental causal factors are identified and classified to produce landslide
susceptibility maps. Given that the earlier landslide inventory had been restricted by only including those landslides
with a verifiable date and location, the decision was made to re-produce the inventory with a greater emphasis on
spatial, rather than temporal, accuracy.
An updated landslide inventory was produced using satellite and airborne technologies. Freely-available, Landsat
Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM+) images were used in conjunction with high
resolution (5m) digital terrain data acquired from Synthetic Aperture Radar (SAR). Both datasets were required so
that both shallow landslides, which are most easily identified through multi-spectral methods (Fig. 4), and deep-
Fig. 4|
False Colour Composite (FCC) images (acquired 19/04/2002) for the Western Province case study domain area with (A) showing a
542 band combination and (B) showing the 457 band combination. The colours associated with different features (landslides, towns) are
illustrated in insets A1, A2, B1 and B2.