基于EnKF排放清单反演方法的关键影响参数评估与优化

以中国一氧化碳(CO)排放反演为例,利用敏感性分析手段评估了集合数目、局地化半径､膨胀因子、观测站点密度和观测数据时间分辨率对排放清单反演的影响.结果表明:站点密度是影响排放反演的最重要参数.在不同站点密度下,反演的中国CO排放总量差异可达34%.同时,站点密度还会影响排放反演对其他参数的敏感性.随着站点密度的降低,排放反演对局地化半径、集合数目和膨胀因子参数变得更为敏感,但对观测数据时间分辨率的敏感性则有所下降.因此在站点稀疏地区,局地化半径是排放反演的主要影响参数,集合数目和膨胀因子次之;但在观测站点密集地区,局地化半径和观测数据时间分辨率是主要的影响参数,而膨胀因子和集合数目的影响相对较小.该研究能够为不同尺度的排放反演开展参数优化提供借鉴.在中国CO排放反演案例(站点密度为1.55个/10⁴km²)中,建议反演参数设置为:集合数目为50、局地化半径为100km、最大似然估计膨胀方案(MLE)、日均或小时观测数据.

A study on evaluation and optimization about key effect parameters of emission inventory inversion method based on EnKF

The ensemble Kalman filter (EnKF) emission inversion is one of the most widely used methods to evaluate air pollutant emission inventory, but its performance is influenced by various parameters. How to identify and optimize the important parameters is the key to ensure the reliability and efficiency of emission inventory inversion. In this study, we use the sensitivity technique to investigate the effects of the number of ensembles, localization radius, inflation factor, observed station density, and temporal resolution of observations on emission inversion for Chinese carbon monoxide (CO) emissions. The results show that the observed station density is the most important parameter affecting emission inversions. The difference in total inversed (CO) emissions in China with varying station densities approaches 34%. Simultaneously, the observed station density also influences the sensitivity of emission inversions to other parameters. As the station density drops, the sensitivity of emission inversion to the localization radius, the number of ensembles and inflation factor increases, while the sensitivity to the temporal resolution of observations diminishes; Therefore, in areas with sparse observations, the localization radius is the most influential inversion parameter, followed by the number of ensembles and the inflation factor; however, in areas with many observed stations, the localization radius and the temporal resolution of observations are the main influential parameters, while the inflation factor and the number of ensembles have relatively less influence. This study can be used to optimize parameters for emission at different scales. The following parameters are proposed for Chinese CO emission inversion (station density is 1.55/10⁴km²): the number of ensembles is 50, the localization radius is 100km, the maximum likelihood estimation (MLE) inflation method, and the daily average or hourly observational data.