| 摘要: | 近年來光學衛星被廣泛用來做大面積的環境監測，其中懸浮沉積物濃度(SSC)變化為眾多監測項目中相當重要的目標之一，在光學影像中懸浮沉積物的濃度越高，通常反映出越高的光譜反射率。有鑑於複雜的環境變因，前人研究大多由衛星影像獲取水體反射率，配合水樣採集的濃度紀錄，建立區域性的預測模型。卑南溪位於台灣東南部，發源於中央山脈，豐沛的颱風降雨與陡峭短促的河道，造就卑南溪成為台灣重要的沉積物輸出河。本研究彙整水利署與環保署長期的懸浮沉積物濃度監測紀錄，以及多顆衛星(FORMOSAT-2、SPOT-4、SPOT-5、SPOT-6)於台灣上空定期拍攝的影像資料。從中篩選2005~2013年53筆同期的實測資料與衛星影像，探討水體反射率與實測濃度的變化關係，並於大氣校正工作後得到最佳的濃度預測模型，接著再以2014年的實測資料進行預測模型的分析與驗證，並發現多元回歸分析有較準確的預測結果，R2為0.9766，預測斜率為1.0431。本研究後段以2011年為例，利用預測模型成功提升懸浮沉積物濃度監控之頻率與擴散分布之細節，整合全年的水文資來分析濃度變化的控制因素，並多次監測出懸浮沉積物濃度已達觸發高濃度重力流之門檻。透過水體反射率標準差分析可能的觀測誤差來源發現，預測結果易受環境影響，特別是河道乾濕季的變化以及海面波浪的干擾，可能都是觀測誤差的主因。若未來能針對此誤差因素進行改善，相信對懸浮沉積物濃度之監測可提供更準確的預測結果，並大幅提升監測頻率與範圍，進而彌補傳統定點人力監測的不足。;Suspended sediment concentration (SSC) is an important indicator of sediment output. Recently, some SSC predictions had been carried out by using optical satellites imagery in different areas. In general, the more suspension sediment in water can directly reflect the higher reflectance of solar radiation. Therefore, most studies developed unique relationships by relating field measurements of SSC to reflectance data from satellite imagery. In this study, we focused on the Pinan River estuary which is born from the largest river in eastern Taiwan. In order to identify an appropriate SSC-reflectance model, we combined our optical satellite images, which included FORMOSAT-2, SPOT-4, SPOT-5 and SPOT-6, with the field data from 2005 to 2013. After doing atmospheric correction, we got the best model with Multiple Regression analysis method. The important thing is that the method has more accurate in predicting SSC, after proving our model with the latest field data in 2014. In the final part, we used the model to resupply the SSC data in 2011, and discussed the characteristics of sediments output with rainfall and discharge. Actually, it is useful for us to replace those stations to get the SSC distribution outside the estuary. And, there are several hyperpycnal flow events occurred at the bottom of the estuary, while the SSC exceeding the threshold (40,000ppm). We also discussed the characteristic of spectral and the source of errors from environment effects. While getting more ways to reduce those noises, we could have better model to predict the SSC. At the same time, it also could enhance the frequency and range of monitoring, and make up for a lack of manual monitoring. |
