黄河源区玛曲土壤湿度模拟不确定性及同化研究

Study of Uncertainty and Assimilation of Soil Moisture at Maqu in the Source Region of Yellow River

在土壤湿度的模拟过程中,由于资料的不足和初始值、驱动数据、参数等的不确定性,影响了模型的模拟精度。本文基于一维土壤水分运动方程,在黄河源区玛曲探讨初始值不确定性对土壤湿度模拟的影响,以及开展土壤湿度同化实验。结果表明:通过模型的预热可以减小不同初始值得到的模型模拟结果的差距;在资料匮乏区域开展土壤湿度研究或者进行区域土壤湿度模拟时,可以采用集合初始值对模型进行预热,将预热期末不同初始值得到的土壤湿度的均值作为预测期的初始值,降低预热期初始值的不确定性;在预测期,采用无迹加权集合卡尔曼滤波UWEnKF可以有效提高土壤湿度的预测精度。因此,结合集合初始值以及UWEnKF既可以降低初始值的不确定性,又有助于改善土壤湿度模拟结果。

In the soil moisture simulation,its accuracy is affected by the few data,and the uncertainty of initial values,forcing data,parameter,etc.This study discussed the influence of initial value uncertainty on soil moisture simulation and conducted the soil moisture assimilation experiment based on one dimensional soil moisture movement equation at Maqu in the source region of Yellow River.The results show that the difference between two soil moisture simulations with different initial values can be reduced through model spinning-up.When conducting soil moisture research at the area with insufficient data or simulating regional soil moisture,the model can be spined-up with an ensemble of initial values,and use the mean value of different soil moisture simulations at spin-up period to be the initial value for the forecasting period and reduces the uncertainty of initial value of spin-up period.The accuracy of soil moisture simulation can be improved by unscented weighted ensemble Kalman filter(UWEnKF)in forecasting period.Overall,it not only reduces the uncertainty of initial value,but also improves the soil moisture simulations by combining the ensemble of initial values and UWEnKF.