岔口小流域非点源污染模型AnnAGNPS不确定性分析

Uncertainty analysis of an AnnAGNPS model used in the Chakou watershed

为了降低模型模拟计算的不确定性,提高模型在北方典型黄土丘陵沟壑区的适用性,以岔口小流域为研究区,选取可能对非点源污染模型——AnnAGNPS产生影响的10个参数,采用修正的摩尔斯(Morris)分类筛选法,分析参数对模型模拟结果影响的敏感性程度。在此基础上,尝试综合利用一阶误差分析法(FOEA)和自助法(Bootstrap)不确定性分析方法识别重要的不确定性参数,评估参数对模型输出变量的不确定性的贡献。综合两者分析结果表明:对径流模拟计算不确定性影响较大的是径流曲线数、田间持水量;对泥沙、氮磷负荷模拟计算不确定性影响较大的是径流曲线数、地形因子、土壤侵蚀因子、耕作管理因子、水土保持因子、化肥施用量。径流曲线数对模型输出结果的不确定性影响最为显著,表明径流产生过程是泥沙和各种营养盐等侵蚀、搬运的驱动力。泥沙和氮磷的不确定性参数较为一致地表明泥沙是氮磷输出的载体,即在短时暴雨事件下氮磷输出形态通常以泥沙结合态为主,泥沙的不确定性直接影响和决定氮磷等营养物输出的准确性。模型输出结果中泥沙不确定性最大,径流不确定性最小。研究表明,流域水文和土壤侵蚀过程的模拟是校准模型、降低不确定性的关键。在控制和治理流域水土流失及非点源污染时应着重考虑径流曲线数、施肥措施、农田管理方式和水土保持情况。

In order to reduce the uncertainty of model simulations and improve the applicability of models in the Chakou watershed of a typical loess hilly-gully region, in this study, we selected ten parameters of an AnnAGNPS model and used the Morris classification method to analyze the sensitivity of the parameters to the simulation results of the model. The uncertainty analysis method FOEA（First-order error analysis）and Bootstrap were adopted to identify important uncertain parameters and evaluate the contribution of these parameters to the uncertainty of the model output variables. The following results were obtained. The number of runoff curves and field water-holding capacity had a considerable influence on the uncertainty of the runoff simulation. Furthermore, the number of runoff curves, terrain factor, soil erodibility factor, tillage management factor, soil and water conservation factor, and chemical fertilizer application amount had considerable influences on the uncertainty of sediment, nitrogen, and phosphorus load calculation. The number of runoff curves had the most significant effect on the uncertainty of the model output, which indicated that the runoff generation process is the driving force of erosion and transportation,affecting sediment and the runoff of various nutrients. The uncertain parameters of sediment and nitrogen and phosphorus were consistent,which indicated that the sediment was the carrier of nitrogen and phosphorus output. In short rainstorm events, the forms of nitrogen and phosphorus output were overwhelmingly dominated by the sediment-bound state, and the uncertainty of sediment directly affected the accuracy of nitrogen and phosphorus output. The uncertainty of the sediment in the model output was the largest and the runoff uncertainty was the smallest. On the basis of these results, we conclude that simulation of hydrological and soil erosion processes is a key factor contributing to model calibration and uncertainty reduction. To control and manage soil erosion and non-point source pollution, the number of runoff curves, fertilizer measures, farmland management, and soil and water conservation should be considered.