密云区是北京重要的地表饮用水源地,准确模拟地表径流量,对于分析泥沙和污染物的运移十分重要.近年来,学者们运用径流曲线数(soil conservation service curve number,SCS-CN)模型计算本区地表径流量,但预报精度不理想;未考虑降雨过程和雨强对于产流过程的影响,可能是造成预报误差的重要原因.该文利用密云石匣小流域5个坡面径流小区共201场降雨产流资料,提出次产流径流曲线数计算方法,以改进SCS-CN模型并分析改进后模型模拟效果.结果表明,次产流径流曲线数与多年平均径流曲线数的比值和最大30 min降雨量与次雨量的比值之间呈显著幂函数递增关系,据此提出计算次产流径流曲线数的幂函数方程,以改进SCS-CN模型.当曲线数为0.02时,改进后模型模拟效果最好,效率系数为0.693,明显高于未改进的SCS-CN模型.改进后模型对裸地和耕地的产流模拟效果较好,但对林地的产流模拟效果不理想.今后需在深入分析产流机理的基础上,进一步提出与土壤特性有关的模型参数优化方法.%Miyun District is in the region for drinking water source for Beijing, which is a mountainous region. Because of steep slopes there, there may be severe soil erosion during storms. To analyze the movement of sediments and nutrients, accurate estimation of surface runoff is important. In recent years, the soil conservation service curve number (SCS-CN) model has been used in the mountainous region of Beijing, but model accuracy is unsatisfactory. In various storm events, the parameter-runoff curve number varied over a wide range. If the influence of rainfall processes and characteristics is not considered, simulation error for the surface runoff can be large. In the present study, data observed for rainfall and runoff depth during 201 rainfall-runoff events from experimental plots with various land cover and management were used to improve the SCS-CN model and test modeling accuracy. The 5 experimental plots were in the Shixia watershed, northeast of Miyun Reservoir, covering117°01'-117°07'E, 47°32'-47°38'N. Observed runoff depth data for 127 rainfall-runoff events were used to improve the SCS-CN model; the other runoff depth data for 74 events were used to test modeling accuracy. Based on analyses of the influence of rainfall processes and intensity on the runoff and curve number for each rainfall event, a method for calculating curve number for each rainfall eventwas proposed. This indicated that the ratio of curve number for each rainfall event to the annual mean curve number increased with the ratio of maximum 30-minute rainfall to total rainfall for the event, with a power function relationship. The power function for curve number for each rainfall event calculation improved SCS-CN modeling accuracy. Nash-Sutcliffe efficiency, correlation coefficient r, and mean relative error (MRE) were used in the examination of simulation results. To achieve optimum modeling accuracy, a range of initial abstraction ratio values from 0.01 to 0.30 was tested for the improved model. An initial abstract ratio 0.02 was used in the improved SCS-CN model so that Nash-Sutcliffe efficiency was 0.693, r was 0.859, and MRE was 4.21%. The Nash-Sutcliffe efficiency for the SCS-CN model without improvement was only 0.151. Because the study area is dominated by a monsoon climate, in the rainy season, storms with relatively high rainfall intensity were common. The ratio for rainfall that infiltrated was smaller, so the initial abstract ratio value was smaller than that in the USA. Simulation results for different antecedent moisture conditions (AMCs) were as follows. For dry conditions, instead of total rainfall amount, rainfall intensity may be more important to the process of infiltration excess runoff. For humid conditions with greater soil moisture contents, rainfall amount may be more important to surface runoff formation. Simulation results for various land uses were different with the Nash-Sutcliffe efficiency of 0.713 and 0.735 for the improved SCS-CN model used for bare land and cropland, respectively. On the surface of the bare land with little vegetation and cropland with low vegetation cover (except for corn stems), from the effect of rainfall splashing and surface runoff scouring, rills formed provided runoff paths. The Nash-Sutcliffe efficiency was low for woodland because the formation mechanism of surface runoff was complex, and excess infiltration-saturation runoff may occur during certain rainfall events. Moreover, interannual variability of vegetation cover for shrubland and woodland may alter the runoff coefficient. In 1994 and 2000, the runoff coefficients for shrubland and woodland were 0.058 and 0.057 respectively; in 2013 and 2015, the runoff coefficients were 0.140 and 0.032 respectively. Optimization of parameters related to soil properties is needed in future research on the improved model.
展开▼
