應用遙測技術於臺灣北部地區水文循環之研究

Study on the Hydrologic Cycle of the Northern Taiwan Using Remote Sensing Techniques

吳治達(Chih-Da Wu)

指導教授 ： 羅漢強 ； 鄭祈全

國立臺灣大學/生物資源暨農學院/森林環境暨資源學研究所/博士(2010年)

https://doi.org/10.6342/NTU.2010.02441

摘要

環境變遷與集水區水文循環之間的關係已成為環境規劃的重要課題。國內外已有許多專家學者結合大氣環流模式 (General Circulation Models, GCMs) 與GWLF河川流量模式 (Generalized Watershed Loading Functions, GWLF)，探討氣候變遷對集水區水資源的衝擊效應。但大部份的研究除受限於大尺度地表蒸發散量的調查不易之外，在蒸發散覆蓋係數 (Evapotranspiration Cover Coefficient, CV) 之設定方面，亦大多根據GWLF手冊中所列之參考值進行設定。然真實地表之土地使用類別甚為複雜，如僅依手冊中之參考值進行參數設定，可能會影響分析結果之正確性。此外，土地使用型態和蒸發散量的逐年變化，亦會影響集水區之未來水文狀態，而傳統的流量模擬研究甚少針對此兩項因子之影響效應加以探討。
　　有鑑於此，本研究以台灣北部地區為試區，旨在利用遙測技術推估真實地表的蒸發散量與CV值，以提昇流量模擬之正確性，進而結合SEBAL模式 (Surface Energy Balance Algorithm for Land, SEBAL)、CGCM1大氣環流模式 (The First Version of the Canadian Global Coupled Model; CGCM1) 與Markov模式，模擬未來土地使用型及蒸發散量的變化，並分析其對未來流量模擬之影響，最後再綜合氣候、土地使用及蒸發散量等環境變遷因子，進一步評估台灣北部地區未來水文循環可能遭受之衝擊效應。研究方法首先利用遙測混合式分類方法進行台灣北部地區之大地資源衛星(Landsat-5)的土地使用分類，並配合數值地形模型(Digital Terrain Model, DTM)與SEBAL能量平衡模式，先計算與蒸發散量有關之16項環境參數，再推估地表蒸發散量，並比較各土地使用型之蒸發散量，過程中為評估空間尺度和生態分類系統二因子對環境參數之影響，乃選定兩種空間尺度 (台灣北部地區及其轄內的7個集水區)和兩種生態分類系統 (12個地理氣候分區及7個集水區)，透過多變量逐步判別分析方法，探討該二因子對環境參數之影響效應；其次，在利用SEBAL模式計算蒸發散量和應用遙測方法推估CV值之後，進而應用GWLF模式模擬淡水河集水區之河川流量，目的除了驗證流量模式之適用性之外，並評估傳統查表方法和遙測方法所推估之CV的流量模擬差異；最後，以兩期土地使用資料為基礎，整合Markov模式與CGCM1模式，預測未來短、中、長期之土地使用變遷、並推估其CV，再經由GWLF之流量分析，分析土地變遷及蒸發散變化對於未來河川流量之影響，進而評估北台灣地區水文系統可能遭遇到的衝擊效應。
　　研究結果指出，北台灣地區經混合式影像分類後，計分為森林、建地、水體、耕作農地、休耕農地、雲及陰影7類土地使用型，其整體分類準確度經檢核區檢定後為89.00%；在土地使用型之蒸發散量方面，以森林最大 (一月：0.723cm；七月：0.395cm)，建地為最小 (一月：0.220cm；七月：0.088cm)；在空間尺度和生態分類系統對環境參數之影響分析方面顯示，使用不同生態分類系統和空間尺度來區分5種土地使用型 (森林、建地、水體、耕作農地及休耕農地) 所需要的環境參數與參數數目皆不盡相同，但常態化差異植生指標與地表熱紅外光放射率兩項參數，不管在那一種生態分類系統，均為重要的判別參數；在利用蒸發散量和土地使用兩因子模擬河川流量之結果顯示，利用遙測推估之CV值(濕季：1.245；乾季：0.851) 與查表所得之CV值 (濕季：0.842; 乾季：0.717) 確實有差異。若透過流量站的觀測資料，並結合

關鍵字

遙測技術 ； 水文循環 ； SEBAL模式 ； GWLF模式 ； CGCM1大氣環流模式 ； 蒸發散覆蓋係數 ； 流量模擬

並列摘要

Watershed hydrology, especially stream flow, is expected to be highly sensitive to the influences of global climate change. Traditional studies have integrated the General Circulation Models (GCMs) with the Generalized Watershed Loading Function (GWLF) model to estimate stream flow rates. However, using these models in the context of a transitioning climate and on a large spatial is problematic, particularly for the estimation of two important parameters, evapotranspiration (ET) and cover coefficient (CV).
This study focuses on an integrated analysis of the hydrological cycle using remote sensing techniques to estimate the ET and the CV. Furthermore, we improved on older studies by integrating the Surface Energy Balance Algorithm for Land (SEBAL) model, the First Version of the Canadian Global Coupled Model (CGCM1), and the Markov model which allows us to predict land-use and ET change. The results were applied to assess the future impacts of global warming on hydrological cycles of northern Taiwan. Our methods include applying hybrid image classification to generate the land-use maps of the northern Taiwan using Landsat-5 images; using digital terrain model (DTM) and the SEBAL model to calculate 16 environmental parameters relevant to ET. We then compared the differences among different land-use types; (1) investigating the effects of two ecosystem classification systems (i.e., watershed division method and geographic climate method) at various spatial scales on environmental parameters using stepwise discriminant analysis; (2) comparing stream flow simulations according to the GWLF model with two CV values derived from remote sensing and traditional methods; (3) integrating the Markov model and the CGCM1 model to predict future land-use and CV parameters for evaluating the effect of land-use change and ET change; and (4) finally, assessing the future impacts on hydrological cycle of the northern Taiwan.
The results indicated that the study area was classified into seven land types (i.e., forest, building, water, farmland, fallow farmland, cloud-covered, and shadow-covered) with 89.09% classification accuracy. These last two land types could not be analyzed further. A comparison of daily ET values among different land-use types revealed differences. In this study, forest ET is the largest (January: 0.723cm; November: 0.395cm) while building is the smallest (January: 0.220cm; November: 0.088cm). These differences contrive to exist for ecosystem classification systems at various scales, but depend on the selected environmental parameters and the number of parameters included in the model. Two parameters, a normalized difference vegetation index and an emissivity are important factors for discriminating land types. On the aspect of land-use and ET effects on hydrological simulations, the stream flows simulated by two estimated CVs were different. The stream flow simulation using the remote sensing approach (wet season: 1.245; dry season: 0.851) presented more accurate hydrological characteristics than the traditional approach (wet season: 0.842; dry season: 0.717). Meanwhile, according to the result of regression analysis, the flow simulation using RSCV (remote sensing based CV; regression coefficient = 0.877) would represent truer flow characteristics than the use of REFCV (reference CV; regression coefficient = 0.853). In the prediction of future land-use and ET, due to the increase of building area from 13.36% in 1995 and 14.05% in 2002 to 38.91% in 2030, 52.13% in 2052, and 62.36% in 2086, the predicated CV values for next three periods display a decreasing trend no matter under which climatic change storyline. In addition, land-use and ET change indeed affect the predicted stream flows. The predicted flows with consideration of these two factors were lower than those without consideration. Finally, the impact assessment on the hydrology of the northern Taiwan indicated that the flow volumes increase due to urban expansion, ET decline, and climate change, and it will lead to the increase of stream flow.
From above results, obviously the integration of remote sensing, the SEBAL model, the CGCM1 model, and the Markov model is a feasible scheme to predict future land-use, ET change, and stream flows. Therefore, it can be extended to the further studies in water resource management and global environmental change.

並列關鍵字

Remote sensing techniques ； Hydrology ； SEBAL model ； GWLF model ； CGCM1 model ； Evapotranspiration cover coefficient ； Stream flow simulation

參考文獻

Cheng CC, Chen, YK, Jan JF, Wang SF (2005) DTM, GIS, and DSS applications in forestland ecosystem classification and suitability analysis. Journal of Photogrammetry and Remote Sensing 10:351-360.

Cheng CC, Lo HC, Chen YK, Wu CD (2008) Assessment of ecosystem classification systems at various spatial scales on environmental parameters using remote sensing techniques. Journal of Photogrammetry and Remote Sensing 241–254.

Cheng CC, Wu CD, Chuang YC (2007) Influence of land-use changes and climate change on stream flow simulations: a case study of the Jiao-Long watershed. Taiwan Journal of Forest Science 22(4):483–495. (in Chinese, with English abstract).

Cheng CC, Wu CD, Wang SF (2005) Application of Markov and Logit Models on Monitoring Landscape Changes. Taiwan J For Sci 20(1):29–36.

Chen CT, Wu ST, Chiang YF (2006) Using MODIS satellite images to estimate evapotranspiration in Taiwan. Taiwan Journal of Forest Science 21(2):249–261