應用概似不確定性估計(GLUE)於集水區營養鹽總最大日負荷規劃

The Application of Generalized Likelihood Uncertainty Estimation (GLUE) to A Nutrient Total Maximum Daily Load (TMDL) Plan

施恆益(Heng-Yi Shih)

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(2013年)

https://doi.org/10.6841/NTUT.2013.00657

摘要

總最大日負荷(Total Maximum Daily Load, TMDL)是目前美國環保署在全國積極推動的集水區污染總量管理之規劃，而總最大日負荷中的安全差距量(Margin of Safety, MOS)，一般皆以主觀且任意的百分率估算，欠缺較嚴謹的學理分析，所以本研究試圖將模式模擬的不確定性分配到安全差距量。
本研究旨在應用SWAT(Soil Water Assessment Tool)模式結合不確定性分析，規劃翡翠水庫營養鹽之總最大日負荷。研究收集1995年至2011年翡翠水庫集水區之氣象、水文與地文資料，應用概似不確定性估計(Generalized Likelihood Uncertainty Estimation, GLUE)結合拉丁高階方塊取樣法(Latin Hypercube Sampling, LHS)進行模式檢定及不確定性分析，並以驗證後的SWAT模擬來推估翡翠水庫集水區長期的汙染負荷。
本研究按照甲類水體水質標準，分別針對翡翠水庫氨氮、硝酸鹽氮和總磷規劃總最大日負荷，並將概似不確定性估計之不確定性範圍納入總最大日負荷規劃中，量化安全差距量。研究結果顯示模擬的水庫硝酸鹽氮、氨氮、總磷安全差距量分別佔TMDL比例為0.47%、5.33%、22.97%。而模式模擬的水庫總磷超過總最大日負荷，需要進行污染負荷削減，所以本研究以模式進一步模擬五種最佳管理作業(Best Management Practices, BMPs)，分別為階段平台(Terrace)、田埂(Field border)、肥料混入(Manure incorporation)、肥料管理(Nutrient Management)與邊坡穩定結構物(Grade Stabilization Structure)，來削減總磷負荷，其中以肥料混入對總磷的削減效率為最好。另外模擬結果顯示五種最佳管理作業組合之複合式最佳管理作業，對總磷削減效率為42.85%，可使翡翠水庫15年年平均總磷模擬最大濃度(0.0173 mg/l)低於總最大日負荷(0.02 mg/l)。
研究結果顯示概似不確定性估計結合拉丁高階方塊取樣法，可被有效地應用於評估不確定性，並量化總最大日負荷中的安全差距量。而研究結果所得的安全差距量，可提供水庫管理決策者，評估翡翠水庫氨氮、硝酸鹽氮和總磷濃度可能的不確定性範圍。並且可根據本研究最佳管理作業模擬的削減效率結果，評估適合的最佳管理作業，來進行非點源污染負荷削減，以確保翡翠水庫水質能符合水質標準。

關鍵字

非點源污染 ； 營養鹽 ； SWAT模式 ； SWAT-CUP ； 不確定性分析 ； 拉丁高階方塊取樣法(LHS) ； 概似不確定性估計(GLUE) ； 總最大日負荷(TMDL)

並列摘要

The Total Maximum Daily Load (TMDL) plan is a main watershed management approach to improve the impaired water quality under the Clean Water Act of USA. In addition, margin of safety (MOS), one of the major components in TMDL implementation, which accounts for uncertainties about the relationship between pollutant loads and receiving water quality, is usually evaluated by a specified percentage instead of more precise quantification. This research aims to scientifically quantify the MOS in TMDL program by employing the uncertainty analysis for model’s predictions.
The SWAT (Soil and Water Assessment Tool) model’s inputs and their uncertainty effects on model simulations for assessing the nutrient loadings in Feitsui Reservoir watershed were fully investigated. The effects of output uncertainty in model input parameters were evaluated by employing Generalized Likelihood Uncertainty Estimation (GLUE) based on Latin hypercube Sampling (LHS). The results of uncertainty analysis revealed that the MOS value of the reservoir nitrate nitrogen, ammonia nitrogen, total phosphorus (TP) loadings are 0.47%, 5.33%, and 22.97% of TMDL amounts, respectively.
Moreover, the validated SWAT model will be employed to evaluate the long-term water quality impacts due to the implementation of several Best Management Practice (BMP) scenarios. In the first place, the results of TMDL calculation show that the simulated average annual TP maximum concentration (0.0349 mg/l) has exceeded the maximum allowable concentration (0.02 mg/l) which was calculated based on the water quality standards of Category A water bodies. Therefore, second BMP scenarios were developed to reduce TP loading and elevate the water quality within standard of Category A.
Various BMP scenarios consisting of terrace, field border, manure incorporation, nutrient management, and grade stabilization structure are further simulated for TP reduction. The simulation results show that manure incorporation is the best BMP scenario. Additionally, the implementation of five BMPs combination scenario will make average annual TP maximum concentration (0.0173 mg/l) to satisfy the water quality standard (0.02 mg/l).
This study concluded that GLUE combines LHS can be effectively applied to assess uncertainty, and quantify the MOS component in a more scientific approach during TMDL assessment. The results of this study demonstrates an effective management approach for administration authority to control pollution efficiently and protect water quality with specified standards for all water bodies.

並列關鍵字

nonpoint source pollution ； nutrient ； SWAT model ； SWAT-CUP ； uncertainty analysis ； LHS ； GLUE ； TMDL ； MOS ； BMPs

參考文獻

47. 吳政緯，翡翠水庫營養鹽之總最大日負荷規劃，碩士論文，國立臺北科技大學土木與防災研究所，臺北，2010。

59. 黃宇齊，翡翠水庫及水庫集水區水文暨水質模擬與其不確定性，碩士論文，國立臺北科技大學土木與防災研究所，臺北，2010。

57. 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，臺北，2009。

60. 廖倫偉，應用概似不確定性估計於溼地參數之不確定性分析，碩士論文，臺灣大學生物環境系統工程學研究所學位元論文，臺北，2011。

56. 連以婷，水文模式之參數不確定性分析，碩士論文，臺灣大學生物環境系統工程學研究所學位元論文，臺北，2010。