HR: 0800h
AN: G51B-0085    [Abstracts]
TI: The Hazard Analysis of Debris Flow by a Neural Network and Digital Terrain Model: A Case Study in Taiwan
AU: Shiau, C
EM: techno@rhce.dyns.cx
AU: Lee, C
EM: ct@gis.geo.ncu.edu.tw
AU: * Lee, C
EM: leecs@mail.ntou.edu.tw
AB: Taiwan is a narrow island. Due to the movement of the plate tectonics, the mountain ranges are almost in N - S direction. On the island, there are a lot of mountains, but few plains. Because of the rapid development on economies, the plains are nearly used-up. So people are trying to develop the mountainous land. However, when the typhoon, heavy rain, or earthquake occurred, there triggers the catastrophic events, such as the landslide and debris flow. Among these catastrophes, the debris flow is the most horrified and it happens frequently year after year. Thus, how to prevent the debris flow and pass this information to warn the local residents becomes an urgent matter. The most reliable method to identify the hazardous debris flow is to investigate in the field. Because the change of the factors, the debris flow may happen. So it will need a long-term monitoring system. The requirement of the manpower and financial resources will increase significantly. Although there are much studies on applying the GIS to the debris flow problem, but it­Ýs still difficult to become a practical tool due to the data gathering and law limitation. If a fast-accurate and self-recognized computer model is available, this will save the time and efforts. In this study of debris flow by using the neutral network and digital terrain model, we select the Chen-Yu-Lan River in Nantou (middle of Taiwan) as an area for study. The debris flow caused in 2001 right after the TORAJI typhoon. Using the GIS software, the MapInfo, which is much simple to get the numerical data from digital terrain model. We have firstly used the 15 factors to start the training. We obtain a successful ratio of about 88%. There are some areas can­Ýt be explained. We believed it is caused by the other factors of debris flow that weren't included. The geoscientific factors are important. However, because of the resolution of present-available geological data don't reach the standard of this study. Therefore, they are skipped. Other relative data, such as the area of tumble, and density of vegetation etc. are not available at the present database.
DE: 1709 Geodesy
DE: 1645 Solid Earth
SC: Geodesy [G]
MN: 2004 AGU Fall Meeting