PM_(2.5)传输模型的初始场优化

Optimization of Initial Field of PM_(2.5) Transport Model

根据PM_(2.5)的传输特征,利用伴随同化技术建立了PM_(2.5)伴随同化模型。将此模型应用于中国大陆地区,对PM_(2.5)初始场进行优化研究。根据中国各大城市在PM_(2.5)污染情况,设置了华北、华东地区呈开口向下的椭圆抛物曲面和西部地区呈开口向上的椭圆抛物曲面这2种模型的初始分布,并利用PM_(2.5)伴随同化模型对其进行优化。实验结果表明,代价函数、观测点处绝均差、优化结果与给定分布的绝均差均有明显下降,优化结果也非常接近给定分布,说明该模型能够合理有效地对初始场进行优化。同时,对2种给定分布的优化情形对比分析猜测,有效观测点个数不足可能是导致优化结果不够理想的原因。为此设计了减少非有效观测点个数和增加有效观测点比例2组敏感性实验,发现有效观测点所占比例越大,优化效果越好。该研究为以后PM_(2.5)的准确模拟打下了基础。

The adjoint assimilation scheme was applied to a PM（2.5） adjoint assimilation model construction, based on the transport characteristics of PM（2.5）. The model was applied to the Chinese mainland, and the PM（2.5） initial field was optimized.The pollution situation of Chinese mainland features two ideal PM（2.5） initial field distributions involving parabolic surface with down convex in North China and East China and upward convex in western China. The forementioned distributions were optimized by the adjoint model. Experiment results suggest that the cost functions, mean absolute errors of observations, and mean absolute errors of optimization results and prescribed distributions were decreased drastically after assimilation. The optimization results were also rather more approximate to the prescribed distributions, indicating that PM（2.5） initial field could be efficiently estimated. Meanwhile, the optimization result might be poorly favorable because of lack of effective observations, compared with optimization process of the two forementioned prescribed distributions. Two sensitive experiments involving increases and decreases effective observations were implemented, and optimization efficient was more obvious if ratio of effective observations was increased. This study prepares for the future accurate PM（2.5） simulation.