Using GIS and Artificial Neural Network for Studying the Variance of Shoreline – A Case of Kenting National Park

• Main Authors: Lian-tsai Deng, 鄧連才
• Other Authors: Yu-min Wang
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/97650765499508564566

碩士 === 國立屏東科技大學 === 土木工程系所 === 100 === In this study, the survey an discussion are focused on the six beaches, Houpihu, Nanwan, Dawan, Shiauwan, Tsunfansu and Shadau, at Kenting National Park between March, 2010 to February, 2011. A tool has been developed by using geographic information system (GIS) for establishing zero meter elevation contour line and exporting fifty coordinate [X, Y] at equal distance for the input of artificial neural network (ANN). The ANN involved in the study was back propagation neural network model which is available in the NeuroSolution software. Based on the data obtained, the relationship between or among beaches were categorized into two scenarios. The results showed typhoon event is the main reason that affect beach volume change in Kenting National Park. However, it is also found the beach volume change will recover on downslope wind season. Through the results of ANN study to compare changes between beaches, scenario 1: coefficient of determination 0.74 from all year round data and 0.87 from rainy season data of Tsunfansu was the lowest while the other beaches were all higher than 0.9. In addition, scenario 2: the results showed when Houpihu data was applied as input, the Tsunfansu coefficient of determination is respectively about 0.85 and 0.9 for rainy season and all year round data. Only using data of downslope wind season, the best model performance was found by applying Houpihu, Nanwan and Dawan as input which the coefficient of determination was higher than 0.9