论文部分内容阅读
陆面数据同化由于能将观测数据和模型模拟有机结合,已逐步发展为地球科学研究的重要方法之一.通过数据同化方法在模型中不断融入新的观测数据,一方面可以有效地校正陆面过程模型的预测轨迹,提高模型状态变量的估算精度,另一方面可以不断减小模型中的不确定因素,优化模型中的相关参数.在众多数据同化算法中,粒子滤波算法不受模型线性和误差高斯分布假设的约束,适用于任意非线性非高斯动态系统,逐渐成为当前数据同化算法研究的热点.本研究基于残差重采样粒子滤波算法发展了一个数据同化方案,将微波亮温数据同化到大尺度半分布式VIC(Variable Infiltration Capacity)陆面水文模型中,对土壤水分进行估算,并对模型中的三个水力参数进行同步优化.最后设计了一系列对比实验并利用美国亚利桑那州在SMEX04(Soil Moisture Experiment 2004)期间获取的一套完整的实验数据对该同化方案进行了验证.结果表明,该同化方案能够大幅度提高土壤水分估算精度,同时模型中的三个水力参数也得到了较好的优化,从而证明了该数据同化方案的有效性.
Land Data Assimilation Because of the combination of observation data and model simulation, it has gradually developed into one of the important methods in earth science research.Using data assimilation method to integrate new observation data into the model, on the one hand, it can effectively correct the land surface Process model to improve the estimation accuracy of the model state variables, on the other hand, the uncertainties in the model can be continuously reduced and the related parameters in the model can be optimized.In many data assimilation algorithms, the particle filtering algorithm is not affected by the model linearity and The constraint of error Gaussian distribution is applicable to any nonlinear non-Gaussian dynamic system and has become a hot research topic in data assimilation.In this paper, a data assimilation scheme was developed based on residual resampling particle filter algorithm, which assimilates microwave brightness temperature data To the large-scale semi-distributed groundwater hydrological model of VIC (Variable Infiltration Capacity), the soil moisture was estimated and the three hydraulic parameters in the model were optimized synchronously.A series of comparative experiments were designed and used by the authors SMEX04 (Soil Moisture Experiment 2004) obtained during a complete set The validation data of the assimilation scheme are validated.The results show that the assimilation scheme can greatly improve the accuracy of soil moisture estimation and the three hydraulic parameters in the model are also well optimized, which proves the effectiveness of the data assimilation scheme Sex.