Please use this identifier to cite or link to this item:
標題: 應用迴歸分析法於最佳水質指標選擇之研究
作者: Chen, Paris Honglay
David, D-W.Tsai
Lu, Ming-Yuan
關鍵字: RPI;RPI;WQI;迴歸分析;集水區管理;WQI;Regression analysis;Watershed management
出版社: 臺中巿: 國立中興大學農學院
Project: 農林學報, Volume 58, Issue 4, Page(s) 307-321.
The main purpose of this study was to find out the best water quality index with simple and multiple regression analysis, which could be helpful for monitoring and decision-making in watershed management. We selected the Dansuie River Watershed as a study site and collected all water quality data available. Regression analysis was applied to find out the regression relationship between water quality indexes (RPI and WQI) and water quality parameters. Also we make a comparison study between RPI and WQI. The results could help us to evaluate these two water quality indexes. We find out several important results in this study: (1) In simple regression, the design of RPI, dissolved oxygen, BOD and ammonia were significant parameters but suspending solid was not. Besides the nutrient pollution, WQI could point out the potential problem of pH. (2) The result of full model selection for multiple regression showed that the most suitable model was formed with parameters selected by simple regression. And (3) After comparing all the regression analyses, WQI exhibited a better performance for water quality. WQI also offered better sensitivity than RPI. Therefore, WQI is a better index for water quality management.

本研究主要目的在於探討最佳水質指標,以協助集水區水質管理策略之訂定與監測。研究中選用淡水河流域為研究區域,在蒐集水質資料後,嘗試使用迴歸分析法,分別分析RPI(River Pollution Index)、WQI (Water Quality Index)兩種水質指標與本區16項水質參數之迴歸關係,藉此了解水質指標之解釋機制,並依據分析結果來評斷水質指標之優缺點。研究結果顯示在單迴歸分析中RPI之組成溶氧、BOD、氨氮均為顯著參數,而懸浮固體則不顯著;而在WQI分析結果則顯示此流域除營養鹽污染外,可能存有水中酸鹼平衡遭破壞的問題。透過不同複迴歸全模式選擇法發現,以單迴歸顯著參數所建構之複迴歸全模式,最符合相關理論,且具有合理之模式解釋能力,RPI模式R^2=0.682,WQI模式R^2=0.88。經過比較RPI與WQI之單迴歸結果後,發現WQI明顯含括範圍較廣,相較於RPI可更完整的解釋水質狀態。而在複迴歸分析方面,結果同樣顯示WQI可解讀較多的水質資訊,因此本研究在綜合所有迴歸分析結果後,明顯指出WQI的敏感度優於RPI,因此建議未來水質管理可將WQI指標納入優先考量。
Appears in Collections:第58卷 第04期

Files in This Item:
File SizeFormat Existing users please Login
151262-5.pdf949.06 kBAdobe PDFThis file is only available in the university internal network   
Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.