利用Landsat TM資料以Mahalanobis影像分類法分析德基水庫之藻華現象

Analysis of Algal Bloom in Techi Reservoir Using Mahalanobis Classification Method with Landsat TM Data

章國威(Kuo-Wei Chang) ； 吳啟南(Chi-Nan Wu) ； 陳伯中(Pei-Chung Chen)

《航測及遙測學刊》 8卷2期 (2003/06) Pp. 13-26

https://doi.org/10.6574/JPRS.2003.8(2).2

摘要

本文以經鏡檢所得之二角多甲藻的數量為地真(Ground Truth)資料，並依此地真資料來選取訓練區(Training Region)，再以Mahalanobis監督式影像分類法(Supervised Image Analysis)，來分析德基水庫二角多甲藻(Peridinium bipes Stein fa. occlatum)之藻華區域。所得之藻華影像分析圖與地真調查之結果相近，其皆顯示藻華發生之區域多位於水庫上游區域，且藻華發生之區域及面積會隨季節變化而有所不同。究其原因發現，水庫上游集水區為梨山溫帶水果及高麗菜的主要產區，大量的營養鹽隨著山坡土壤的沖蝕而流入水庫上游，因而提供二角多甲藻大量繁殖之良好環境。此外，由於蔬果之施肥期及其施肥量會隨作物之生長期的不同而有所不同;故造成沖蝕流入水庫之營養鹽量會隨著季節的不同而有差異，因此使得藻華發生之區域面積會隨季節變化而有所不同。以監督式影像分類法分析德基水庫之藻華區域，不但可以同時獲得整體之水體資料，叉可以同一分析方法分析不同時期之水體資料，來消除可能之人為誤差，故此法不失為藻華監測之有利工具。

關鍵字

德基水庫 ； 二角多甲藻 ； 藻華 ； 遙測 ； Mahalanobis監督式影像分類

並列摘要

We used the observed dinoflagellate densities as ground truth data and selected the training regions with the ground truth data then analyzed the dinoflagellate (Peridinium bipes) algal bloom areas in the Techi reservoir using mahalanobis classification method. The results of Mahalanobis classified images and ground truth data are similar. They all showed that algal bloom areas most located on the upstream of the Techi reservoir. The algal bloom sites and areas varied with seasons because many agricultural activities on the upstream of Techi reservoir and fertilizers ran off to the reservoir which provided a good environment for the growth of Peridinium bipes. The reason why the algal bloom site and areas varied with seasons is because that the fertilized period and fertilized quantities varied with the different growth period of vegetation, and as the results, fertilized quantities ran off to the reservoir varied with different 'seasons. In conclusion, Mahalanobis classification method is a powerful tool for monitoring algal blooms, by getting entire water body data and temporal water body data simultaneously.

並列關鍵字

Techi Reservoir ； Dinoflagellate ； Algal Bloom ； Remote Sensing ； Mahalanobis Classification Method