水資源在世界各國都屬於重要資源之一，乾淨的水源更是民生、工業、農業等不可或缺的資源，需要加強管理與監控。雖然台灣屬於水源豐沛之地區，但儲水並不容易，且近年來受到極端氣候影響，台灣也面臨了高強度暴雨與長時間乾旱的問題，因此對於水庫的水質管理更顯重要。有鑑於台灣目前監測站都屬固定點測量為主，不足以代表整個集水區，且現有的實地測量皆以一個月進行一次測量，對於整個集水區短時間內詳細的變化難以掌控。本研究使用水文模式(SWAT)與衛星影像(SPOT與Landsat)，探討石門水庫集水區上游與庫區內的水質濁度監測問題，首先先使用SWAT模式對石門水庫集水區進行模擬，獲得每日一次的懸浮沉積物濃度，對於石門水庫集水區有長時間並進一步的理解。再使用衛星影像建立石門水庫集水區預測模型，並進行校準與驗證，才能使用衛星影像進行水質濁度預測。使用SWAT模式進行篩選，找出濁度上升事件，再使用衛星影像之預測模型進行預測，討論其準確度與濁度上升之問題；另於2015年石門水庫發生長達數月之乾旱，針對其時段進行衛星影像之預測。根據結果衛星影像應用於石門水庫庫區可行度較高，相對誤差平均約30%，水庫庫區的水質預測較為準確，主因是水庫庫區內面積範圍大，不易受到周圍沙洲或植被所影響。上游集水區則受限於水深，若在乾旱時期進行監測，容易因水位高度不足1m而產生誤差，導致獲得之反射率過高，誤判水質呈現混濁狀態。;In facing the extreme weather, Taiwan requires adequate water quality management. The current monitoring stations in watersheds around Taiwan are fixed point measurements, not enough to represent the entire catchment area. And currently the frequency of data gathering of these monitoring stations is less than once a month, which makes the number of samples are very limited. In this study, hydrological model (SWAT) and satellite imagery (SPOT and Landsat) were used to study the turbidity of water in the upstream and reservoir areas of Shimen Reservoir channel water surface area. The prediction model of turbidity of Shimen reservoir water surface area was established by satellite imagery, and calibration and verification were carried out. SWAT was employed to identify high turbidity events and then the satellite image was utilized to estimate turbidities. Discussion of the accuracy of turbidity estimation model is provided. According to the results, the application of satellite images to estimate the turbidity of Shimen reservoir water surface area is highly feasible, with the relative error being about 30%. The turbidity estimation in the water surface of reservoir is more accurate than that in the channel water surface area. This is mainly because the area of the reservoir area is open, free from being affected by the surrounding sandbar or vegetation. The accuracy of upstream channel water surface area is limited, for the depth of the water is shallow and would be affected by the noises due to high reflection. Furthermore, if the monitoring is carried out during the drought period, it is likely to be error-prone, due to the height of the water level being less than 1 m which resulting high reflectivity and resulting misjudging high water turbidity.
