集合Kalman滤波在土壤湿度同化中的应用

Applications of Ensemble Kalman Filter in Soil Moisture Assimilation

基于非饱和土壤水模型和集合卡尔曼滤波（Ensemble Kalman Filter,简称EnKF）并结合陆面水文模型——可变下渗能力模型（Variable Infiltration Capacity,简称VIC模型）发展了一个土壤湿度同化方案。利用1998年6～8月淮河流域能量和水循环试验（HUBEX）项目外场观测试验区——史灌河流域梅山站土壤湿度逐日观测资料及1986～1993年合肥和南阳两站点的土壤湿度旬观测资料进行同化试验,结果表明该同化方案能完整估计土壤湿度廓线,同化的土壤湿度与观测资料基本吻合,反映了土壤湿度的日、旬、月、季变化,同化方案是合理的。与基于扩展卡尔曼滤波（Extended Kalman Filter,简称EKF）的土壤湿度同化方案的结果比较,基于EnKF的土壤湿度同化方案易于实现,且通过选择恰当的集合样本数其同化效果总体上略优于EKF同化方案,但前者同化时需要花费较多的计算时间。

A soil moisture assimilation scheme is developed, which is based on the ensemble Kalman filter （EnKF）, an unsaturated soil water flow model and the Variable Infiltration Capacity （VIC） land surface hydrologic model. To validate and verify the scheme, several assimilation experiments with in-situ observations and infiltration derived from the VIC model driven by observed meteorologic forcings are presented. The results demonstrate that the assimilation scheme can retrieve the true soil moisture profile, improve the estimation of soil moisture, and capture the daily, dekad, monthly and seasonal variation of soil moisture, which shows that the scheme is reasonable. Comparisons of the assimilation experiments with the algorithms EnKF and Extended Kalrnan Filter （EKF） show that the EnKF scheme with the suitable ensemble size is easy to implement and is a little superior to the EKF scheme for soil moisture assimilation, but the computational cost for the EnKF scheme is higher than that for the EKF scheme.