伏牛山区陶湾流域径流泥沙模拟误差分析

以豫西伏牛山区陶湾流域为研究对象，利用该流域2006—2007年22次径流、泥沙数据对分布式AnnAGNPS（Annualized AGricultural NonPoint Source）模型进行校准，2008—2009年31次径流、泥沙数据对模型进行验证。选取R²（决定系数）、E（效率系数）、VE（误差比）等统计参量评估流域径流、泥沙、氮、磷物质输出的模拟精度。结果表明:（1）SCS-CN值是径流模拟精度的主要影响因素，校准期内径流VE=-7.7%（R²=0.95，p < 0.05），验证期内VE=-6.1%（R²=0.90，p < 0.05），误差值位于可接受的范围之内， < ±15%。径流误差相对较低，径流模拟精度也影响泥沙和氮、磷的模拟精度；（2）泥沙模拟误差影响因素较为复杂，植被覆盖率、曼宁粗糙系数等因素最为敏感，还受流域DEM、土壤、土地利用等空间参数精度的影响。校准期内VE=15.1%（R²=0.55，p < 0.05），验证期内VE=17.0%（R²=0.60，p < 0.05），泥沙模拟误差较径流要大。泥沙和径流模拟误差趋势相同，对小型降雨事件，模拟值偏高，大型降雨事件模拟值偏低。（3）氮、磷模拟值具有更大的不确定性，氮VE=22.0%（R²=0.69，p < 0.05），磷VE=24.0%（R²=0.48，p < 0.05）。AnnAGNPS模型对径流、泥沙、氮、磷模拟均存在有一定的不确定性，模拟误差呈现一定的变化趋势。校准后的AnnAGNPS模型可以用于伏牛山区流域。

Analysis of Simulation Error of Runoff and Sediment for Taowan Watershed in Funiushan Mountain Area

A typical watershed, Taowan, was selected as the study site, and the main goal was to evaluate the performance of AnnAGNPS Model in simulating runoff, sediment loading and nutrient loadings under the conditions of Funiushan Mountain area, and analyze the influencing factors of the simulation error. The data of 22 events of runoff and sediment in the watershed during the period from 2006 to 2007 were used to calibrate the model and the data of 31 events during the period from 2008 to 2009 were used for validation purposes. The whole evaluation consisted of determining the coefficient of determination (R²), Nash-Sutcliffe coefficient of efficiency (E), and the percentage volume error (VE). Most of model input parameters were sourced from Luoyang Forestry Bureau field observations and experiments. Results showed that: (1) runoff was undersimulated by -7.7% with R² of 0.95 (p < 0.05) during calibration period (2006—2007) and undersimulated by -6.1% with R² of 0.90 (p < 0.05) during validation period (2008—2009), factors affecting runoff simulation errors were accuracy of values for the SCS-CN, the results of runoff simulation also affected the other objects, which may be the reason of relatively low error of runoffs, the model simulated runoff within the range of acceptable accuracy (< ±15%); (2) the model oversimulated the event-based sediments by 15.1% with R² of 0.55 (p < 0.05) during calibration period (2006—2007) and 17.0% oversimulated with R² of 0.60 (p < 0.05) during validation (2008—2009), factors affecting sediment simulation error are more complex. Vegetation coverage and Manning roughness coefficient were most sensitive to the sediment simulation errors. And it was also affected by the scales of spatial parameter accuracy such as watershed DEM, soil and land uses. The trend of sediment simulation errors was similar to runoff. For the events of small magnitude, the model generally oversimulated them, while the opposite was true for larger events. Nitrogen was oversimulated by 22.0% with R²=0.69 (p < 0.05), and phosphorus was oversimulated by 24.0% with R²=0.48 (p < 0.05). In general, the model performs well in simulating runoff compared to sediment and nutrients. As a watershed management tool, it can be used under the conditions of Funiushan Mountain area after proper calibration.