image, only landslides greater than about 625 m2 (2×2 pixels) can be
mapped accurately.
The steep topography of the study area makes for thin layers of soil
and colluvium with thicknesses of b2 m. The field investigation of
landslides shows that the depth of failure in most of the landslide
areas exceeds 3 m. A prevailing landslide type in this area is a complex
type composed of soil and rock failures.
Besides, in order to understand the change of landslides in
different events, we calculate the landslide ratio (total landslide
area to the whole catchment area), newly-formed ratio (newly formed
landslide area to the total landslide area after a given event),
and reactivated ratio (reactivated landslide area to the landslide area
that existed before a given event) by comparing the landslide map
with the maps of earlier landslide events (Lin et al., 2008a; Chuang
et al., 2009). Newly-formed landslides and reactivated landslides can
be surveyed by contrasting the post-event satellite images with the
pre-event images. In this study, the time spans between pre-event
and post-event satellite images are less than two months. A newly formed
landslide will not have appeared in the pre-event satellite
image, which means that the area was not affected by landsliding in
the two months before the typhoon events. In contrast, reactivated
landslide refers to an area that had been affected by a prior typhoon
event, which was again disturbed in the subsequent typhoon event
(Lin et al., 2008a).