Application of Artificial Neural Network and SatelliteImages to Water Quality Estimation – A Case Study of Shihman Reservoir

• Main Authors: Chi-Hsien Chang, 張騏顯
• Other Authors: 陳正炎
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/hk42fx

碩士 === 國立中興大學 === 土木工程學系所 === 97 === The Eutrophic Status of a reservoir results from the plankton reproduction due to an increase of the nutrition in the reservoir. Total phosphorous, Secchi depth, chlorophyll-a are the three typical parameters affecting the eutrophic status. In situ investigation is the traditional approach for acquiring the data of the three parameters, whereas it is time-consuming and inefficient. Moreover, an analyzed result of the point-basis data usually can not be used to well assess the water quality of the whole reservoir. Currently, the satellite remote sensing technology with in situ investigation has been widely applied to environmental monitoring, so is also adopted in this paper. The Shihman Reservoir, which is the largest reservoir in northern Taiwan, is selected to be the study site of this research. The experimental materials in this paper include 30 frames of SPOT series satellite images and the three parameters investigated in situ from May 2006 to November 2008. By gathering the Time Series method-Autoregressive Process of Order p, the Neural Network (NN) models of the spectrums of satellite imagery vs. the eutrophic parameters was developed. A Back-Propagation Network (BPN) model was built to transfer image pixels to eutrophic parameters for the assessment of the Shihman Reservoir.