Wang Shuhuai, Tong Juxiu, Huang Chen
MOE Key Laboratory of Groundwater Circulation and Environmental Evolution, China University of Geosciences (Beijing), Beijing, 100083, People's Republic of China.
School of Water Resources and Environment, China University of Geosciences (Beijing), Beijing, 100083, People's Republic of China.
Sci Rep. 2025 Jul 10;15(1):24929. doi: 10.1038/s41598-025-09858-3.
Nitrogen loss in paddy fields has been widely recognized as a significant contributor to nonpoint source pollution. Predicting this process through modeling is crucial, yet model parameters always carry uncertainty. Clarifying the time-dependent importance of the model parameters can help to better know the process effect such as precipitation or chemical reaction on nitrogen loss. Therefore, to rank the parameter importance, a global sensitivity analysis (GSA) named Sobol method is applied to a nitrogen loss model in paddy fields based on the soil mixing layer theory via modifying HYDRUS-1D model. To reduce the computational cost, local sensitivity analysis (LSA) is applied to the prediction model firstly, and three important model parameters, including soil mixing layer depth (d), soil detachability coefficient (α) and precipitation intensity (p), are selected. Then, the Sobol method is applied to the prediction model to analyze the sensitivities of these three parameters. It is novel but reasonable that the Sobol sensitivity indices (including the first, second and total order indices, FOI, SOI and TOI) of d, α and p vary with time. The study results indicate that the importance of parameter varies with time during a rainfall. In surface runoff, α is most important at early times, while p becomes most important at later times for predicted urea and NO-N concentrations. α is always the most sensitive parameter for the predicted NH-N concentration in surface runoff. In soil, the GSA results are opposite for α and p. Generally, d is less important than α and p, and the interaction between each two parameters reflected by the SOIs has limited importance. d presented sensitivity in LSA but insensitivity in GSA. Sensitivity of α showed similar results in LSA (elasticity) and GSA (TOI and FOI), which decreases in surface runoff and increases in soil. All elasticities of p increase at first and decrease gradually later, while the GSA results of p vary oppositely in surface runoff and soil.
稻田中的氮素损失已被广泛认为是面源污染的一个重要因素。通过建模预测这一过程至关重要，但模型参数总是存在不确定性。阐明模型参数随时间的重要性有助于更好地了解诸如降水或化学反应等过程对氮素损失的影响。因此，为了对参数重要性进行排序，基于土壤混合层理论，通过修改HYDRUS - 1D模型，将一种名为Sobol方法的全局敏感性分析（GSA）应用于稻田氮素损失模型。为了降低计算成本，首先将局部敏感性分析（LSA）应用于预测模型，并选择了三个重要的模型参数，包括土壤混合层深度（d）、土壤可蚀性系数（α）和降水强度（p）。然后，将Sobol方法应用于预测模型，以分析这三个参数的敏感性。d、α和p的Sobol敏感性指数（包括一阶、二阶和总阶指数，即FOI、SOI和TOI）随时间变化，这既新颖又合理。研究结果表明，在降雨过程中参数的重要性随时间而变化。在地表径流中，对于预测的尿素和NO - N浓度，α在早期最为重要，而p在后期变得最为重要。对于预测的地表径流中NH - N浓度，α始终是最敏感的参数。在土壤中，α和p的GSA结果相反。一般来说，d的重要性低于α和p，由SOI反映的任意两个参数之间的相互作用重要性有限。d在LSA中表现出敏感性，但在GSA中不敏感。α的敏感性在LSA（弹性）和GSA（TOI和FOI）中显示出相似的结果，在地表径流中降低，在土壤中增加。p的所有弹性先增加后逐渐降低，而p的GSA结果在地表径流和土壤中变化相反。
