基于AnnAGNPS模型的农业非点源污染模拟

以大沽河典型小流域为研究区，借助GIS和相关资料率定参数，完成模型数据库的建立。在此基础上，采用流域出口2001～2002年径流量、泥沙和总氮监测数据优化参数、校准和验证模型，分析AnnAGNPS模型在大沽河典型小流域应用的可行性。2000～2001年，地表径流的年均模拟偏差分别为12.7%和7%，丰水期的月模拟偏差都在±30%以内，日模拟值与实测值相关系数分别是0.98和0.91，模拟结果理想；年均泥沙输出模拟误差分别为19.1％和15%，月均泥沙估算的相关系数分别为0.78和0.73，模拟结果基本可接受；年均模拟误差为50%左右，月均估计及降雨场次模拟的相关系数分别低于0.7和0.6，总氮污染负荷估算有很大的不确定性。结果表明，如何提高模型的营养盐模拟精度将是未来AnnAGNPS模型在大沽河流域应用的前提与基础。研究成果对于AnnAGNPS模型在国内的应用和推广，具有很好的示范性，并为大沽河全流域分布式水文模型的构建奠定了基础。

Simulation of agricultural non-point pollution based on AnnAGNPS model

GIS(Geographic Information System) was integrated with AnnAGNPS(Annualized Agricultural non-Point Source Model), which was used to input, analyze, and visualize spatial data that agricultural non-point source pollution concerned. On this basis, the performance of AnnAGNPS, used in typical sub-watersheds of Dagu catchment was evaluated. The model was calibrated and validated using observed stream flow, sediment load and total nitrogen data during 2000 and 2001 in sub-watersheds outlet. The annually relative error of surface runoff was separately 12.7% and 7%, the difference during the wet months were in range of ±30%, and R² of 0.98 and 0.91 was separately for event-based runoff, which showed the model performed well for runoff simulation. Moderate accuracy of prediction for sediment was obtained with annually relative error of 19.1% and 15%, and R² of 0.78 and 0.73 during calibration and validation. As the annually relative error of 50% for annual output of total nitrogen, however, AnnAGNPS failed to calculate the total nitrogen with R² separately down to 0.7 and 0.6 for monthly simulation and number of rainfall. It is observed that there is a need to modify or improve the estimation methods of the total nitrogen to increase the performance of the model which can aid watershed management in local conditions. The results in this paper make a good demonstration for applying AnnAGNPS model in China, and lay a good foundation for establishment of whole Dagu basin distributed hydrological model.