探討SWAT模式應用於台灣水田水文模擬之適宜性

Discussion on the Suitability of SWAT Model Applied to Hydrological Simulation of Paddy Field in Taiwan

黃俊壹(Chun-Yi Huang)

指導教授 ： 江莉琦

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202200393

摘要

桃園市為北台灣第二大水稻生產區，其中，南崁溪流域為重要水稻灌區之一，緻密的灌溉渠道網絡提供該地區穩定的灌溉水源。目前國內在水田灌溉管理操作之研究中，尚未有利用水文模式評估水田灌溉管理策略對整體流域之水資源影響，而在使用模式模擬研究進行前，應優先評估模式應用於研究區域是否適用。因此，本研究將利用SWAT水文模式對南崁溪流域進行水文模擬，透過比較灌溉排水操作、評估水田水文特性模擬結果、計算水田之水平衡等，以了解水文模式SWAT模擬水田之水文機制，並評估模式應用於臺灣兩期作水田水文模擬之適宜性分析。
本研究設定研究區域為南崁溪主流及上游茄苳溪、大檜溪流域，研究時間為2009年至2013年兩期稻作。本研究使用SWAT 2012模式，並以壺洞(Pothole)設定來模擬水田型態。為模擬真實水田灌溉面積，本研究以2015 年第二次國土調查圖資為基礎，對照實際水田灌溉面積，建立整合土地使用分類之判斷流程。模式中所設定之灌溉水量，採用農田水利署桃園、石門管理處提供之灌溉統計資料，經彙整所屬小組之不同耕作期灌溉水量，設定計畫灌溉水量，測試模式之壺洞機制，包括:灌溉/排水情境(測試項目一)、壺洞犁底層設定(測試項目二)。SWAT模式水文演算法主要為壺洞產水量之水平衡觀點，透過模式理論水文公式可計算理論壺洞產水量，並比對模式模擬結果，以判斷是否達到水平衡，後續檢視模式反映水田水文變化趨勢之準確度。透過設定之「傳統灌溉」與「自動灌溉」等灌溉方法，分析水文特性變化。比較SWAT壺洞模式及SWAT-Paddy模式之模擬結果，評估模式演算合理性，並對其缺陷提出改善建議。
研究結果顯示，透過土地使用分類整合，模式模擬之水田面積可達約70 %計畫灌溉面積，且空間相關性良好(R2 = 0.82)、均方根誤差(RMSE)為25.46公頃。在測試項目一:多種灌溉/排水情境之河川流量模擬結果顯示，透過蓄/排水操作可模擬水田水量之變化，且可回推研究區域內之實際灌排情況，但在設定排水日流量明顯高估。在測試項目二:壺洞犁底層設定之河川流量模擬結果顯示，本研究建立「以土壤深度作為犁底層深度」與「固定6,000 mm犁底層深度」兩種情境進行模擬，結果顯示以土壤深度作為犁底層深度之情境可有效蓄水，完整呈現水田之水文特性，但在固定6,000 mm犁底層深度之情境因為具有強大土壤入滲作用，使壺洞蓄水量偏少，其模擬水田水文特性之代表性較差。分析壺洞水平衡之模擬時，利用比較理論計算產水量與模式模擬產水量結果，於固定6,000 mm犁底層深度之情境，其分析結果相符，證實水平衡理論成立。但在以土壤深度作為犁底層深度之情境，可以觀察到壺洞模式模擬結果未能完整讀取壺洞產生之地表逕流補注至河川的水量，導致低估理論計算產水量；且理論計算產水量與模式模擬產水量之差值與降雨量有相似趨勢，顯示壺洞模式在降雨逕流關係有修正之必要性。在流量模擬方面，SWAT模式之傳統灌溉及自動灌溉的初始模擬為滿意的，其R2分別為0.64及0.78，模式經參數率定後，傳統灌溉(率定期間: R2=0.7；驗證期間: R2=0.79)與自動灌溉(率定期間: R2=0.82；驗證期間: R2=0.85)均有不錯的模擬表現。然而，SWAT-Paddy模式模擬之流量結果，有出現河川流量突增之異常現象，經多次測試及查證，推測可能因不同SWAT模式版本內部之流量演算法差異所導致。在「傳統灌溉」與「自動灌溉」的模擬結果中，傳統灌溉的產水量普遍較高，而自動灌溉的壺洞蓄水量相對較大，反應傳統灌溉會因為人為灌溉、排水操作產生明顯水文變化。綜觀SWAT模式運用壺洞模擬水田之結果分析，顯示模式能呈現水田特殊水文特性，但對於細部水文演算法仍缺乏精準參數設定及模式理論。因此，在使用 SWAT 模式進行模擬時，應參考實際灌區之灌溉水源及排水設施、優化農業操作參數之設定，並配合實際水田灌溉管理操作與期程，才可更合理模擬流域之水田水文。針對後續SWAT模式改善，本研究建議分析水田水平衡時，應改良模式在壺洞地表逕流補注河川流量之輸出結果，以提高水文模式水田水文模擬之適宜性；SWAT-Paddy模式方面，應整合不同版本之流量演算法，比對流量發生異常之條件並加以改善。

關鍵字

SWAT模式 ； 南崁溪流域 ； 水平衡 ； 壺洞蓄水量 ； 傳統灌溉 ； 自動灌溉

並列摘要

Taoyuan City is the second largest rice producing area in north Taiwan, where its Nankan River basin is one of the important rice irrigation areas with dense network of irrigation channels to provide stable irrigation water sources in the region. At present, in the studies of agricultural management of paddy fields in Taiwan, there is no use of hydrological model to evaluate the impact of strategical management of water resources in the entire basin of paddy fields and thus, in order to use model simulation, it is imperative to evaluate the applicability of such model in the research area. Therefore, in this study, the SWAT model of the Nankan river basin is used to compare the irrigation and drainage in the area, from which the post-simulation assessment would help to calculate the water balance of the paddy fields in the region and evaluate the suitability of applying the SWAT model to the two-stage hydrological processes of the paddy fields in Taiwan.
In this study, the areas of interest included the main stream of Nankan River and the upstream basins of Jiadong River and Dakwai River, and the study time was two periods of rice cultivation from 2009 to 2013. This research would adopt the 2012 SWAT model with a setting of Pothole to simulate the types of paddy fields. In order to simulate the real irrigation area of paddy fields, we would use the second national land survey in 2015 and compared with the actual irrigated area to establish a procedure of integrated land use classification. The quantity of water for irrigation set in the model was obtained from the statistical information of both the Taoyuan and Shihmen Management Offices on the amount of water used during different ploughing seasons of each irrigation team, also including the target depth of irrigation and the testing by pothole, such as: Irrigation/drainage scenario (test item 1) and the condition of plowsole of pothole (test item 2). The hydrological algorithm of SWAT model was primarily based on the balance of water yield in potholes that the theoretical water yield of pothole could be calculated from the theoretical hydrological formula of the model. The simulation results of the model were compared to determine if the water balance was reached, from which the accuracy of the model reflecting the trend of hydrological change in paddy fields could be examined. The change of hydrological characteristics was analyzed between the “traditional irrigation approach” and “auto-irrigation approach”. The simulation results of SWAT-pothole model and SWAT-paddy model were compared and the accuracy of the model’s algorithm was evaluated to provide some suggestions for improving its weaknesses.
The results showed that through an integrated land use classification, the area of paddy fields simulated by the model could cover approximately 70% of the actual irrigated area, with good spatial correlation (R2 = 0.82) and a root-mean-square error (RMSE) of 25.46 ha. For test item 1, the model simulation results of various irrigation/drainage scenarios showed that the change in water volume of paddy fields could be simulated by draining and collecting water, to deduce the actual condition of irrigation and drainage in the study area. However, the model overestimated the daily drainage flow. For test item 2, the simulation results of river flow at the plowsole of potholes under the first scenario of “using the soil thickness as the depth of plowsole” and the second scenario of “fixing the depth of plowsole at 6000mm” showed that scenario one could effectively store water to present a complete hydrological characteristics of paddy fields, while in scenario two, due to the strong soil infiltration of water, there was less water stored in the potholes, showing a poorer presentation of simulated hydrological characteristics of paddy fields. When analyzing the simulation of pothole’s water balance by comparing theoretical water yield with the simulated water yield, the analytical results were consistent, proving that the theory of water balance was valid in scenario two. However, in scenario one, it was found that the simulation of potholes was not able to capture fully the quantity of water for replenishing the rivers by the surface runoff in the potholes, which led to an underestimation of the theoretical water yield. Moreover, the difference between the theoretically calculated water yield and the simulated water yield was similar to that of rainfall, which indicated the need to address the relationship between rainfall runoff in the pothole model. In terms of flow simulation, the initial simulation of traditional irrigation and auto-irrigation in SWAT mode was satisfactory and their R2 were 0.64 and 0.78, respectively. After the model was calibrated with parameters, for traditional irrigation (calibration time: R2= 0.7; validation time: R2= 0.79) and auto-irrigation (calibration time: R2 = 0.82; validation time: R2 = 0.85), they all showed good simulation performance. However, the flow simulation results of SWAT-Paddy model had some anomalies of stream overflow. After many tests and verifications, it was speculated that this could be caused by the differences of flow algorithms within different versions of SWAT model. In the simulation results of "traditional irrigation" and "auto-irrigation", the water yield of traditional irrigation was generally higher, while the water storage capacity of pothole in auto-irrigation was relatively larger, which reflected that traditional irrigation would exhibit significantly observable hydrological changes, affected by manual irrigation and drainage of paddy fields. The results of simulating paddy fields with pothole in SWAT model showed that the model could clearly illustrate the unique hydrological features of paddy fields, despite the lack of accurate parameter setting and model theory for detailed hydrological algorithm. Therefore, when using the SWAT model for simulation, we should refer to the irrigation and drainage facilities of the actual irrigated area and optimize the parameters of agricultural management in accordance with the actual operation and its duration to plan a more reasonable simulation of the hydrology of paddy fields in this study. To further improve the SWAT model, it is suggested that when analyzing the water balance of paddy fields, the output of the model in the surface runoff of the pothole to recharge the river flow should be improved to enhance the suitability of the hydrological model for simulation on the paddy fields. In the SWAT-paddy model, different algorithms of river flow should be incorporated to improve the calculation based on the conditions of abnormal flow.

並列關鍵字

SWAT model ； Nankan river basin ； Water balance ； Water storage capacity of pothole ； Traditional irrigation ； Auto-irrigation

參考文獻

Abbaspour, K. C., Rouholahnejad, E., Vaghefi, S., Srinivasan, R., Yang, H., Kløve, B. (2015). “A continental-scale hydrology and water quality model for Europe: Calibration and uncertainty of a high-resolution large-scale SWAT model”. Journal of Hydrology, 524, 733-752.

Ahn, S. R., Kim, S. J. (2018). “Analysis of water balance by surface–groundwater interaction using the SWAT model for the Han River basin, South Korea”. Paddy and Water Environment, 16(3), 543-560.

Anthofer, J. (2004). “The Potential of the System of Rice Intensification (SRI) for Poverty Reduction in Cambodia”. In: Conference on International Agricultural Research for Developement, Berlin.

Chapagain, T., Yamaji, E. (2009). “The effects of irrigation method, age of seedling and spacing on crop performance, productivity and water-wise rice production in Japan”. Paddy and Water Environment, 8(1), 81-90.

Dai, J. F., Cui, Y. L., Cai, X. L., Brown, L. C., Shang, Y. H. (2016). “Influence of water management on the water cycle in a small watershed irrigation system based on a distributed hydrologic model”. Agricultural Water Management, 174, 52-60.