翡翠水庫及水庫集水區水文暨水質模擬與其不確定性

Modeling Hydrologic and Water Quality Response of FeiTsui Reservoir Watershed and Associated Uncertainty

黃宇齊(Yu-Chi Huang)

指導教授 ： 朱子偉

國立臺北科技大學/工程學院/土木與防災研究所/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00560

摘要

台灣地區河川坡陡流急，又因降雨不均勻，而必須興建水庫以調節有限的水資源供應各種標的用水，然而多數水庫卻面臨因為水庫上游集水區不當管理所造成的水庫優養化問題。翡翠水庫係供應大台北地區民生用水單一目標水庫，水庫水質均受到有關單位的適當保護，但近年來的水質監測卻發現逐漸呈優養化傾向，這其中主要是來自於農業活動與北宜高施工和通車後所帶來的非點源污染所造成。
本研究旨在應用SWAT(Soil and Water Assessment Tool)模式模擬翡翠水庫及其集水區在不同土地利用或管理措施改變所造成的水文與水質衝擊，並探討模式參數之不確定性對模擬結果所造成的不確定性。由總磷濃度機率分佈評估水庫水質呈現優養化之風險，以便於日後研究進一步發現污染關鍵區域，藉由設置適當的最佳管理作業(BMPs)，訂定水庫集水區總最大日負荷(TMDL)計劃，以維護庫區的水質。
SWAT模式係由美國農業部之農業研究中心所研發，此模式整合了CREAMS、GLEAMS和EPIC等模式，為集水區尺度、連續時間模擬的模式，以日為基本時間單位模擬水文、泥砂、作物生長、營養鹽的循環及農藥的傳輸，評估集水區在不同土地利用與土壤分佈的管理作業下對水文、泥砂、營養鹽及農藥承載之影響。
研究中收集翡翠水庫集水區1998至2007年共10年氣象、地文及水文資料，以坪林站的實測值進行模式的檢定和驗證；其中以1998至2002年資料執行模式自動檢定，並以2003至2007年的資料驗證模擬結果。集水區月流量經檢定和驗證後，效率係數分別為0.97和0.54，另外月泥砂量檢定和驗證之後的效率係數分別為0.91和0.33，而月總氮檢定和驗證之效率係數分別為0.57和0.66(年總氮的效率係數為0.83)，月總磷效率係數則分別0.41和0.15(年總磷的效率係數為0.48)。水庫的模擬結果則顯示水庫的月入流量檢定和驗證之效率係數皆為0.97；另一方面在水庫放流中的月總氮在檢定和驗證之效率係數分別為0.93和0.81(年總氮的效率係數為0.95)，月總磷則分別為0.84和0.18(年總磷的效率係數為0.26)。綜合以上結果顯示SWAT模式在對翡翠水庫及其集水區長期的水文與水質模擬均相當合理。可知SWAT模式具有模擬混合土地使用集水區長期水文暨水質反應的能力，進一步應用此模式可有效地管理土地利用與開發及控制非點源污染，期能維護翡翠水庫的水質和水資源永續利用的目標。
水文與水質模式已廣泛地與不確定性分析做結合，並應用於水位、水質及各項水利防災設施之安全性、可靠度及風險評估，以進一步規劃適當的管理措施。為了解輸入參數之不確定性對輸出參數之影響結果，便需進行不確定性分析，研究中考慮各輸入參數之不確定性，包括與土地利用和土壤分佈相關之逕流係數、地表逕流積延係數、土壤飽和水力傳導係數、和其它土壤及地下水相關等，共19項對水文、水質敏感且具不確定性之輸入參數以予變動，進一步分析輸入參數之不確定性及其對模式輸出結果之影響，運用蒙地卡羅（MCS）結合拉丁高階方塊法（LHS）進行100次模擬，以求得各輸出參數之機率分佈，檢視其對水文、泥砂及營養鹽輸出之變化情形，並由卡爾森優養指數法（CTSI）和世界經濟發展組織（OECD）檢定標準法之優養化分級，將總磷濃度機率分佈應用於水庫水質優養化風險評估，結果顯示由世界經濟發展組織的檢定標準法所推估之水質安全可靠度為85%較卡爾森優養指數法93%低，而優養風險為15%則相對較高，期結果有助於水庫優養化的防治策略。不確定分析突顯了參數之不確定性對模擬結果的影響，研究所得各輸出結果的機率分佈可提供決策者更詳細的資訊，進一步評估非點源污染防治方法的可靠度與水質惡化風險，並應用於最佳管理作業(BMPs)與總最大日負荷(TMDL)規劃上。

關鍵字

SWAT模式 ； 非點源污染 ； 營養鹽 ； 不確定性分析 ； 拉丁高階方塊法

並列摘要

Due to the non-uniformly distributed precipitation and the steep-sloped rivers, reservoirs have been widely constructed to manage the water supplies in Taiwan. However, the eutrophication problem that resulted from the poor soil and water conservations as well as inappropriate managements in upstream watershed is prevailing in most reservoirs. The FeiTsui reservoir has been providing raw water for the Great Taipei Metropolitan area, and its water quality has been well protected by associated authorities. Nevertheless, recent monitoring data and researches suggest that the water quality have reached the near-eutrophic state, especially when Taipei-Ilan expressway was built and came to operate. The eutrophication problem is directly correlated to both excessive amounts of nutrient input and heavy rainfalls in the upstream watershed.
This study aims to evaluate the hydrologic and water quality impacts resulting from mixed land uses and various management scenarios on FeiTsui reservoir and its watershed by employing the SWAT (Soil and Water Assessment Toll) model. This study also consider the uncertainty of input parameters, and assessment the eutrophic level of water quality by employ TP concentration probability distribution. Furthermore, the validated SWAT model is also able to assess the long-term water quality impacts due to a complete watershed implementation of BMPs, which will assist the ultimate Total Maximum Daily load (TMDL) development.
SWAT is a river basin, or watershed scale model developed by the USDA-ARS. Specific models that contributed significantly to the development of SWAT were CREAMS, GLEAMS, and EPIC. SWAT model was developed to predict the impact of land management practices on water, sediment and agricultural chemical yields in large complex watersheds with varying soils, land use and management conditions over long periods of time.
Ten-year data (1998~2007) of weather, topography, hydrology, and water quality have been collected for model calibration (1998~2002) and validation (2003~2007). Moreover, the model calibration is performed by using the integrated auto-calibration function. The simulation results of watershed indicate that the efficiency coefficients of monthly streamflow, sediment, total nitrogen (TN), and total phosphorus (TP) for both calibration and validation periods are 0.97/0.54, 0.91/0.33, 0.57/0.66, and 0.42/0.16, respectively. In addition, efficiency coefficients of reservoir monthly inflow, outflow TN, and outflow TP are 0.97/0.96, 0.93/0.79, and 0.85/0.17 for both calibration and validation, respectively. Overall, it is concluded that SWAT’s capacity of modeling hydrologic and water quality impacts from various managements makes it a useful tool to evaluate appropriate watershed development and BMP efficiency for effective control of nonpoint source pollution.
Hydrologic and water quality models are frequently linked with uncertainty analysis, and applied on the safety, reliability, and risk assessment of water flow level, water quality, and relative hydraulic structures, to planning and designing a proper management. The uncertainty analysis of input parameters and its effect on model simulation is also investigated in this study. The variables includes SCS runoff curve number which relates land covers and soil types, surface runoff lag coefficient, soil saturated hydraulic conductivity, and other input parameters which relating to soil and groundwater. Amount 19 sensitive and uncertain parameters ware discussed the influence of model output. The effects of uncertainty in model input parameters will be evaluated by using constrained Monte Carlo simulations based on Latin hypercube sampling (LHS) scheme. Each probability distributions for model output of interest results from 100 model simulations. The developed probability distributions will certainly provide information with desirable probability for specific model output. Finally, both the Carlson Trophic State Index (CTSI) method and Organization for Economic Cooperation and Development (OECD) method were employed to the result of TP concentration probability distribution, to assess the eutrophic level of water quality. The OECD single variable index method result showed the reliability of safety water quality (85%) is lower than the CTSI method result (93%), and the risk of eutrophication (15%) is relative higher. The result of uncertainty analysis highlight the importance of considering input parameter uncertainty when analyzing the outputs of a hydrologic and water quality model. Moreover the resulting probability distributions of model outputs will provide more information for decision-maker in developing the BMPs and TMDL program, and also assessing the reliability of NPS pollution control strategies and risk of water quality degradation.

並列關鍵字

SWAT model ； cnonpoint source pollution ； nutrient ； uncertainty analysis ； Latin hypercube sampling

參考文獻

[41] 陳立宗，翡翠水庫集水區水文暨水質模擬，碩士論文，國立臺北科技大學土木與防災研究所，台北，2009。

[50] 劉柏廷，牡丹水庫集水區非點源污染模擬及風險分析，碩士論文，國立臺灣大學土木工程學研究所，台北，2004。

[53] 謝政道，水庫集水區水土資源管理決策之研究，博士論文，國立臺灣大學環境工程學研究所，台北，2007。

[1] Arnold, J.G., N. Fohrer, "SWAT2000: Current capabilities and research opportunities in applied watershed modelling.", Hydrological Processes 19 (3), 2005, pp. 563-572.

[2] Arnold, J.G., R. Srinivasan, R. S. Muttiah, J. R. Williams, "Large area hydrologic modelling and assessment. Part I. Model development." Journal of American Water Resources Association 34 (1), 1998, pp. 73–89.