二维平流扩散模型下的集合卡尔曼滤波模拟同化研究

Assimilation Research with Ensemble Kalman Filter Using Two-dimensional Advection Diffusion Model

为提高环境数值预报水平,构建了一个针对污染物扩散的模拟数据同化系统。采用集合卡尔曼滤波方法对二维平流扩散模型的状态变量进行了实时校正,实现污染物浓度的实时模拟预报,完成了敏感性实验中集合数目变化、观测方差变化和同化窗口长度变化研究。比较考察观测点位置与污染源距离不同时的预报效果,探讨了优化条件下的同化策略,提出一种根据距离远近动态调节卡尔曼增益权重的方法。在集合数目较小时,可降低计算代价,得到优化的同化效果。

A simulation data assimilation system aiming at pollutant concentration was built for further improving environmental numerical predication level. Ensemble Kalman filter had been applied to real-time adjust state variables of the two-dimensional advection diffusion model,which could simulate and forecast pollutant concentration. Sensitivity tests were conducted with the change of ensemble numbers,observation variance and assimilation window＇s length. Comparing the forecast effects by the different distances between the locations of observation point with pollution source,the assimilation strategies under optimum conditions were discussed,and then a novel method to adjust Kalman gain weight by distances was presented. In addition,the proposed methods could reduce computational cost and obtain better assimilation effect with the smaller ensemble sizes.