【摘要】 NAM模型在国内外流域降雨径流模拟中得到广泛应用，参数敏感性分析是模型构建与应用的重要环节，其目的在于定性或定量评估模型参数对模拟结果的影响，确定参数重要程度，识别敏感参数，以提高模型参数率定的效率。以小清河黄台桥断面以上321 km~2流域为例，利用拉丁超立方抽样方法对输入参数进行随机抽样，以此为基础，分别采用偏秩相关法和互信息法，对NAM模型中9个主要参数进行全局敏感性分析，并对两种方法的结果进行对比分析。结果表明，两种方法得到的参数敏感性次序具有一致性，可以相互验证；对洪峰流量影响最大的参数是地表径流临界值和汇流时间常数，对峰现时间影响最大的参数是汇流时间常数，对径流总量影响最大的参数是地表径流临界值。地下径流临界值、壤中流临界值、壤中流系数、地表蓄水层最大含水量和浅层蓄水层最大含水量为不敏感参数，在模型率定时可以根据经验取固定值以提高率定效率。更多还原

【Abstract】 NAM model has been widely used in rainfall runoff simulation ofwatershed at home and abroad and parameter sensitivity analysis is an important part of model building and application, its purpose is to qualitatively or quantitatively evaluate the influence of model parameters to the simulation results, determine the importance of parameters and identify sensitive parameters, so as to improve the efficiency of model parameter calibration. Taking the 321 km~2 river basin above Huangtaiqiao section of Xiaoqing River as an example, this paper used Latin hypercube method to sample the input parameters randomly. On this basis, partial rank correlation method and mutual information method were used to analyze the global sensitivity of nine main parameters in the NAM model and the results of the two methods were compared and analyzed. The results show that the sensitivity order of parameters obtained by the two methods is consistent and can be verified mutually; the parameters that have the greatest impact on peak runoff are the critical value of surface runoff and the constant of confluence time; the parameters that have the greatest impact on peak time are the constant of confluence time; and the parameters that have the greatest impact on total runoff are the critical value of surface runoff. The critical value of underground runoff, the critical value of soil flow, the coefficient of soil flow, the maximum water content of surface aquifer and the maximum water content of shallow aquifer are insensitive parameters. Fixed values can be selected according to experience in model calibration to increase calibration efficiency.更多还原