对于流域水文模型，获得反映所模拟流域水文循环特征的模型参数是模型应用的重要前提条件。缺资料流域水文模型的参数率定是水文学科的一个研究重点和难点。除传统的参数区域化方法以外，考虑到在部分缺资料流域可能存在少量短时间序列或零星不连续径流量观测数据的实际情况，近年来采用少量径流量观测数据率定模型参数的方法逐渐引起关注。但是以往研究主要针对集总式概念型模型开展。由于参数自动率定将消耗大量时间，当前针对应用较为广泛的分布式模型的研究较少。本研究以我国黑河流域和晋江流域两个气候及产流特征差异性较大的流域分别作为干旱地区和湿润地区的代表流域，研究应用较广的分布式水文模型SWAT(Soil and Water Assessment Tool) 参数率定所需最低连续径流观测的数量，以此探讨使用少量径流量数据校正分布式水文模型的可能性，从而为解决缺资料地区水文模型参数校正这一科学问题提供一种新的可能性。本研究基于采用不同数量的径流量数据对模型进行校正的结果之间的对比开展研究，为了尽量降低参数自动校正方法本身对参数校正的影响，选择通用似然不确定性分析法GLUE（Generalized Likelihood Uncertainty Analysis）作为参数自动率定及模拟不确定性分析方法。从模型模拟效果及模拟不确定性角度对不同数量观测数据的参数率定结果进行对比。研究结果显示：（1）使用一年连续日观测数据，在某些情况下仅使用一个月的径流量率定参数获得参数值的模拟效果与使用多年日观测数据率定所获得的模拟效果近似；（2）使用一整年日径流数据校正模型时，使用丰水年的径流量数据可能比使用枯水年的所获得的模拟效果更好；（3）使用少于一整年的日径流数据校正模型时，在数据量相同的情况下，使用丰水期数据比使用枯水期数据所获得的模拟效果更好；（4）在干旱地区校正分布式水文模型时需要的径流量数据量相较湿润流域更多。本研究的成果对缺资料流域基于水文模型的水资源管理、洪水预报、气候变化及人类活动影响分析研究具有重要的理论价值和现实意义。

Obtaining the model parameters that reflect the characteristics of the hydrological cycle in the simulated basin is an important precondition for hydrological models. Hydrological model calibration in ungauged basin is a hot research in hydrology subject. In addition to the traditional method of parameter regionalization, using a small amount of flow observation to calibrate hydrological models has gradually caused for concern in recent years due to the fact that there actually some short time series or sporadic discontinuous runoff observation data in some ungauged basins. However, previous studies mainly focused on lumped conceptual model. Because the automatic calibration of parameters always consumes a lot of time, current research on distributed model is less used.In this study, there are two large river basins that have great differences in climate and runoff characteristic，Heihe river and Jinjiang river in China，as the representative of the arid and humid regions respectively. Study the required minimum number of continuous runoff observation when calibrate parameters of SWAT (Soil and Water Assessment Tool) a widely used distributed hydrological model in two basins, to explore the possibility of using a small amount of runoff data calibrate distributed hydrological model and provide a new possibility to solve the problem of hydrological model calibration in ungauged basin. Based on the comparison among the results of model calibration using different amounts of runoff data, in order to reduce the influence from parameter automatic calibration method as possible, this study chooses GLUE (Generalized Likelihood Uncertainty Analysis) as the method for automatic calibration of parameters and the uncertainty analysis of simulate results. At last, compare the results that using different amounts of runoff data to calibrate model parameters from angles of model simulation effect and uncertainty. Results show: (1) the simulation results of parameters calibration that use a year of continuous daily observation data, or in some cases, use only a month of runoff data is similar to use many years of daily observation data; (2) when use a whole year of daily runoff data to validate the model, the simulation result using the runoff data in wet years may be better than the result obtained in dry years; (3) when use the daily runoff data less than one year to validate the model, the simulation result using the same amount of runoff data in wet season is better than the result obtained in dry season; (4) the required number of runoff observation for the distributed hydrological model validation in the arid regions is more than humid regions. The results of this study have important theoretical value and practical significance for the research about water management, flood forecasting, climate change and impact analysis of human activities based on hydrological models in ungauged basins.