研析濁水溪流域地表-地下水文與營養鹽之時空分布:SWAT-MODFLOW模式之應用

Abstract

濁水溪流域和濁水溪沖積扇位於台灣中部，是重要的水資源區，涵蓋山地、丘陵和平原，對農業和經濟發展至關重要，隨著經濟發展和人類活動增加，水資源需求和污染負荷增大，工業廢水和生活污水排放，導致營養鹽進入水體，可能污染地表和地下水，這些污染物通過地表逕流和入滲進入地下水系統，對水質管理構成挑戰，水文過程中的地表逕流、河川和地下水交換以及污染物的傳輸和分布有著重要影響。 本研究針對地表水和地下水在水文循環中之互動，特別是降雨產生之地表逕流如何影響地下水系統進行研究，為了全面考量降雨所產生之地面與河川逕流以及降雨入滲機制，整合了SWAT模式和MODFLOW模式，並進一步加入RT3D模式來模擬在濁水溪流域內不同土地利用類型下，營養鹽中NO3-N與PO4-P之傳輸和化學反應。SWAT模式與MODFLOW模式之模擬期間是皆從2000年至2021年，其中在SWAT模式中暖機期為2年(2000-2001年)，採用日模擬，MODFLOW模式以日為單位，時間間距採用月做切分，由於在耦合模式中，SWAT模式有設定暖機期，並不會給MODFLOW模式資料，因此MODFLOW模式的檢定從2002年開始。 研究主要目標包括：(1) 建立SWAT-MODFLOW模式，模擬推估濁水溪沖積扇之降雨補注量、河川與含水層交換量；(2) 評估 SWAT-MODFLOW-RT3D模式之適用性，通過模式率定與驗證，確保模式能夠反應流域內之水文過程；(3) 了解不同土地利用類型下NO3-N與PO4-P在地表水和地下水中之運輸，系統分析其遷移、轉化和最終歸宿。研究成果探討包括SWAT流量模擬結果、MODFLOW率定驗證結果、河川與含水層交換量和NO3-N與PO4-P傳輸結果等。 研究結果顯示，SWAT模式在9個測站中流量模擬表現，根據Moriasi et al.(2015)標準統計值大多滿意(satisfactory)，其中，率定期R2平均在0.71，NSE平均在0.55，且PBIAS大部分在45%以下，而驗證期R2平均在0.67，NSE平均在0.63，PBIAS大部分在45%左右。經率定後，單純的MODFLOW模式的水頭誤差大多在2.5 m以內，年平均抽水量為26.48億噸，年平均補注量為27.23億噸。而耦合後的SWAT-MODFLOW模式，SWAT模式推估之年平均補注量約14.77億噸，年平均補注量約為年平均降雨量0.36倍。河川與含水層的交換率方面，在扇頂與扇尾區域主要是河川入滲至含水層，而在扇央區域則主要是含水層出滲至河川。SWAT-MODFLOW-RT3D 模式結果顯示，NO3-N 和PO4-P濃度隨時間和空間顯著變化，PO4-P的濃度兩者呈現相似的動態，相較於 NO3-N，PO4-P因為容易被土壤吸附和固化，其濃度遠低於 NO3-N，而NO3-N 具有較高的溶解度和移動性，因此濃度較高。 本研究提供了詳細之SWAT-MODFLOW-RT3D模式架構建置方法，可作為其他流域進行水文與環境研究時之模擬方法，具有應用潛力，建議未來在研究上可加強模式改進、長期監測及跨部門協作，以推動水資源的可持續利用與保護，確保區域經濟與生態環境的協調發展，尤其是在氣候變遷的背景下，研究應進一步探討極端天氣事件對水資源和水質的影響，並納入更多的環境變量和人類活動影響，以提升模型的預測能力和適用性，將能更有效地應對未來的挑戰，保障水資源的可持續管理與利用。

The Zhuoshui River Basin and Zhuoshui River alluvial fan, located in central Taiwan, are crucial water resource areas. The basin encompasses mountainous, hilly, and plain, with diverse topography and abundant water resources, essential for agricultural production and economic development. With economic development and increased human activities, the demand and utilization of water resources have imposed a significant load. The demand for water resources and pollution loads have increased, leading to the discharge of industrial wastewater and domestic sewage. This discharge introduces nutrients such as nitrogen and phosphorus into water bodies, potentially contaminating surface and groundwater. These pollutants enter the groundwater system through surface runoff and infiltration, posing significant challenges for water quality management. The hydrological processes of surface runoff, river flow, groundwater exchange, and the transport and distribution of pollutants are critical factors affecting water quality management. This study focuses on the interaction between surface water and groundwater in the hydrological cycle, particularly how surface runoff generated by rainfall affects the groundwater system. To comprehensively consider surface and river runoff and the infiltration mechanisms caused by rainfall, the study integrates the SWAT and MODFLOW models, further incorporating the RT3D model to simulate the transport and chemical reactions of NO3-N and soluble phosphorus under different land use types within the Zhuoshui River Basin. The simulation period for both the SWAT and MODFLOW models spans from 2000 to 2021, with a warm-up period of 2 years (2000-2001) for the SWAT model, using daily simulations. The MODFLOW model operates using daily units, with a time step of monthly intervals. Since the SWAT model has a 2-year warm-up period (2000-2001) that does not provide data to the MODFLOW model, the calibration period for the MODFLOW model begins in 2002. The main objectives include the following. The first objective is to establish the SWAT-MODFLOW model to estimate rainfall recharge and river-aquifer exchange in the Zhuoshui River alluvial fan. The second objective is to evaluate the applicability of the SWAT-MODFLOW-RT3D model by calibrating and validating the model to ensure it accurately reflects hydrological processes within the study area. The third objective is to understand the transport of NO3-N and dissolved phosphorus(PO4-P) in surface and groundwater under different land use types, systematically analyzing their dynamic movement, transformation, and ultimate fate. The results indicate that the flow simulation performance of the SWAT model at nine stations is generally satisfactory based on the statistical criteria established by Moriasi et al. (2015). During the calibration period, the mean R² was 0.71, the mean NSE was 0.55, and most PBIAS values were below 45%. In the validation period, the mean R² was 0.67, the mean NSE was 0.63, and most PBIAS values were approximately 45%. The groundwater head error in the MODFLOW model was mostly within 2.5 meters. The annual average pumping volume was 2.65 billion tons, and the annual average recharge volume was 2.72 billion tons. The SWAT model estimated an average annual recharge volume of approximately 1.47 billion cubic meters, which is about 0.36 times the average annual rainfall. Regarding river-aquifer exchange rates, infiltration from the river to the aquifer was the main process in the fan top and tail regions, while exfiltration from the aquifer to the river was the primary process in the fan middle region The SWAT-MODFLOW-RT3D model results show significant temporal and spatial variations in NO3-N and PO4-P concentrations, with PO4-P exhibiting similar dynamics. Compared to NO3-N, PO4-P is more readily adsorbed and solidified by the soil, resulting in lower concentrations, whereas NO3-N, due to its higher solubility and mobility, exhibits higher concentrations. This study provides detailed methods for constructing the SWAT-MODFLOW-RT3D framework, serving as a reference for hydrological and environmental research in other watersheds and demonstrating potential for application. It is recommended that future research focus on model improvements, long-term monitoring, and interdepartmental collaboration to promote sustainable use and protection of water resources, ensuring the coordinated development of regional economies and ecological environments. Especially in climate change, further investigation into the impact of extreme weather events on water resources and quality is needed. Incorporating more environmental variables and human activity impacts will enhance the model's predictive capabilities and applicability. Through these efforts, we can more effectively address future challenges and ensure water resources' sustainable management and utilization.