Useful assessment of landslide hazards requires, at the minimum, an understanding of both ‘where’ and
‘when’ that landslides may occur. As Fig. 1 shows, landslides result from a combination of factors, which
according to (Dai and Lee, 2002) can be broadly classified into two categories: (1) preparatory variables
that make the land surface susceptible to failure without triggering it, such as slope, soil properties,
elevation, aspect, land cover, and lithology; and (2) the triggering events that induce mass movement, such
as heavy rainfall and glacier outburst. For rainfall-triggered landslides, at least two conditions must be met:
the areas must be susceptible to failure under certain saturated conditions, and the rainfall intensity and
duration must be sufficient to saturate the ground to a sufficient depth. Therefore, to diagnose the landslide
occurrence, the proposed system must link two major components: landslide susceptibility (LS)
information and real-time precipitation analysis, as shown in Fig. 1. The LS map empirically shows part of
the “where” and the rain intensity-duration primarily determines the “when” information. In use, the
“where” LS map is overlaid with real-time satellite-based rainfall “when” layer to detect landslide hazards
as a function of time and location.