The Comparative Study of Statistical Outlier Tests on Water Monitoring Data of Major Rivers in Taiwan

• Main Authors: Gordon K.S. Lin, 林高陞
• Other Authors: 林正芳
• Format: Others
• Language: en_US
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/81304515179605609978

碩士 === 國立臺灣大學 === 環境工程學研究所 === 91 === Outlier identification is the first step to ensure the quality of data. Environmental monitoring data containing outliers have impacts on performances of its use. The research not only introduces outlier identification by quantitative errors, logic determination, and exceeding legal limits, but by means of statistical methods. The research uses water quality monitoring data of seven monitoring stations on major rivers in Taiwan, i.e. Dan Shuei River 1003, Shin Dian Stream 1019, Shin Dian Stream 1018, Shin Dian Stream 1272, Kao Ping Stream 1200, Kao Ping Stream 1202 and Er Ren Stream 1190. It uses Analysis-it + General 1.67 in conjunction with Microsoft Excel to help graphing and analyzing. Four statistical outlier tests; 1.5 IQR test, 95th percentile test, Rosner test and Walsh test, are carried out on the water monitoring data. Results are analyzed by considering four major indicators; BOD, COD, SS and NH3/N, and water quality standards of surface water body; Jia, Yi, Bing and Ding. Each test identified distinct suspected outliers for different indicators and water bodies. The results are interesting and promising as various statistical outlier tests can be applied for intended purposes according to their characteristics found on the study.