| Summary: | 碩士 === 國立中興大學 === 水土保持學系所 === 100 === Typhoon Morakot caused massive collapses in the watersheds which resulted in debris disasters and channel blockages of the nearby villages especially located at the southern Taiwan. Due to scattered distribution of the landslides, how to extract the watershed landslides, find the landslides occur reasons, and monitor the status of watershed’s vegetative restoration can be served as the references of delineating potential landslide areas, assessing the vegetative recovery rate (VRR) and providing the proprity management of the landslides.
This study extracted the spatial distribution of landslides and calculated the collapse ratio in Cishan Creek watershed using satellite images before and after the Typhoon event. Besides, the VRR of the watershed was evaluated by applying the image taken a year after the hits of Morakot. The results show that the collapse ratio and VRR of the watershed were 8.07% and 7.24% respectively. The topographic factors of the watershed such as watershed area, watershed perimeter, watershed length, length of mainstream, rivers of total length, number of rivers, elevation, relief, number of headwaters, slope, watershed width, form factor, stream frequency and stream order are employed to study the contributions to the collapse ratio and VRR of the watershed. Among them, the factors of watershed perimeter, number of headwaters, number of rivers, elevation, slope, relief, watershed width and form factor can be classified as categories of scale, gradient, and form three principal component axes, which extract about 83.79% of information and show significant effects on the collapse ratio and VRR of the watershed.
The VRR can further be grouped as excellent, ordinary, and poor rates, the collapse ratio be grouped as high, medium and low by K-means, which couples with discriminant analysis can derive Fisher''s linear discriminant function to explore the affecting factors of the collapse ratio and VRR for the watershed landslides. The results indicate that the accuracy of the classification for the collapse ratio and VRR can reach 67.5% and 70%. The topographic factors of a watershed can effectively interpret the potential landslides and VRR. The difference of categories derived from K-means and discriminant analysis can be used as the index of management priority. No. 21, 22, 28, 31, 33及36 are the management compartments which needed to be more concerned in Cishan Creek watershed, and the models developed in this study can be as the references of watershed landslide management.
|
|---|
