A study with artificial neural network on estimating rainfall by using satellite cloud image

• Main Authors: Tain-De Sun, 孫天德
• Other Authors: An-Pei Wang
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/37342336885135940578

碩士 === 中原大學 === 土木工程研究所 === 94 === There are debris flow disasters in recently years in Taiwan. Since present warning system is hard to reach satisfied results, debris flows frequently cause serious loss on not only human’s life but also their property. Heavy rain is one of the factors that caused debris flows. The precipitation usually concentrates during typhoon season from May to October every year. A question as to how to estimate typhoon rainfall rapidly and accurately has become very important for early warning of debris flows. The purpose of this study is to learn about estimating typhoon rainfall by using Artificial Neural Network (ANN) with cloud temperature. The Shihmen Reservoir and its watershed are taken as example area of study,and data of cloud temperature and rainfall of invading typhoon from 1996 to 2003 are collected in this study. A temperature-rainfall model is established to predict rainfall at 3 hours later in Shihmen Reservoir. Two results are found: Firstly, rain stations with similar geographic properties have similar temperature-rainfall models. It means that landform affects rainfall condition. The past references also demonstrate this. Secondly, for the same rain station, the cloud temperature relates highly with heavy rainfall. The results display that the model performed well especially in the big typhoon events. Although for the small typhoon events the model did not perform as good as for the big ones, the errors are still acceptable. The results of this paper could serve as a fine reference for predicting debris flow induced by typhoon invasion.