淺層土壤水動態模型之建立與應用

土壤水分動態在自然界中扮演十分重要的角色，如地表與地中的能量交換等，易受雨量、氣溫及地文等條件影響，對森林及集水區管理而言，是個重要的考量因子。然而，相較其他地面水文的資料，臺灣土壤水的現地資料相對短缺。為能有效評估土壤水分動態，本研究欲建立一淺層土壤水系統動態模型，藉由物理機制模擬不同土壤質地下不同深度的淺層土壤水含水率，以及入滲、滲漏、蒸發、飽和流及漫地流等水文過程。於模型建立後，本研究於玉峰集水區設置四個樣區，用以校正與驗證本模型，並驗證此模型的可行性與精確度。研究結果發現利用物理機制所建立而成的淺層土壤水動態模型，以玉峰集水區現地土壤水資料做校正與驗證，可得到擬合度佳且具可信度的小時土壤水模擬。該地區的淺層土壤水的實測值介於1.50 %～35.40 %，模擬值介於3.00 %～43.40 %，模擬結果的平均誤差（ME）介於－2.33～2.38，平均絕對誤差（MAE）介於1.08～6.24，均方根誤差（RMSE）介於1.37～8.15，平均絕對百分比偏差（MAPE）介於28.31 %～57.78 %。依據敏感度分析結果顯示，雨量、含石率、孔隙形狀參數、飽和含水率、田間含水率及細切土層數等皆為土壤含水率及各水文過程變化的敏感因子。本研究新建立的淺層土壤水動態模型，利用當地氣候觀測資料及土壤性質特性，模擬土壤水含量的小時變化，可協助補足土壤水資料不足的情況。此外，因本模型係以物理機制為基礎，能夠適用於多數場域，並具有分析未來氣候變遷情境的潛力，對於農業、災害等議題，可提供相關水文過程及土壤水變動等資訊，作為決策參考。

關鍵字

土壤含水率；物理模型；系統動態；水文過程；土壤含水率分佈；敏感度分析

並列摘要

The soil moisture dynamics plays a key role in natural processes, such as heat exchange between surface and subsurface, and response to the magnitude of precipitation, air temperature and landscape conditions. As such, the soil moisture has been considered as an important factor for forest and watershed management. However, compared to surface hydrological measurements, soil moisture information is very scarce in many watershed areas in Taiwan. To provide soil moisture estimations, I constructed a physically–based model considering multiple processes, including infiltration, percolation, evaporation, saturation and overland flow, to simulate soil moisture at various depths in different soil textures. Modeling results were calibrated and validated with field measurements from four sampling sites in the Yufeng watershed, to examine the feasibility and accuracy of the model. Calibration and validation results from the Yufeng watershed showed that the dynamic shallow soil moisture model performed well in providing reasonable well-fitted and reliable simulations of soil water contents in an hourly time frame. The field measurements of the soil water contents were between 1.50 % to 35.40 %, while the simulation results ranged between 3.00 % to 43.40 %. Modeling mean error (ME) ranged between －2.33 to 2.38, mean absolute error (MAE) between 1.08 to 6.24, root mean square error (RMSE) between 1.37 to 8.15, and mean absolute percentage error (MAPE) between 28.31 % to 57.78 %. Based on the sensitivity analysis, precipitation, soil–gravel mixture content, shape parameter, saturated water content, field water capacity and divided layers appeared critical factors to influence the soil moisture dynamics and the associated hydrological processes. In sum, this newly constructed dynamic shallow soil moisture model uses meteorological data and soil information to simulate hourly change of soil water contents, and can be applied to assist the provision of required soil water information that were often insufficient in many places. As the model is based on physical mechanisms, it can be generally applied to other regions. More importantly, it has the ability to provide forecasts dealing with agricultural or natural disaster issues in future climate change scenarios to find mitigation strategies, and to offer science–based information for decision making copping with potential changes in hydrological processes and soil moistures.