A Study On The Application Of Ensemble Optimal Interpolation In GRAPES＿Meso5.1/CUACE

Atmospheric chemical transport model（CTM）is a key tool to carry out fog-haze forecasting and warning,however,CTM has unavoidable errors.The data assimilation（DA）technique is one of the important methods to reduce the uncertainty of CTM inputs and improve the model forecast accuracy.In this paper,based on the regional online chemical weather numerical forecasting system（GRAPES＿Meso5.1/CUACE）,the influence of the selection of key parameters of the ensemble optimal interpolation（En OI）DA method on the assimilation effect is analyzed using sensitivity tests,and the improvement effect of PM_2.5 on the analysis and forecast fields by the En OI assimilation of automatic stations is quantitatively evaluated using forecast-assimilation cycle tests.Based on this,an assimilation module for relative humidity was developed and a preliminary analysis of the effect of this module on relative humidity,PM_2.5 and visibility forecasts were performed.The main conclusions are as follows:（1）The assimilation of PM_2.5 greatly reduces the uncertainty of the initial PM_2.5field.Both the mean error and root mean square error（RMSE）of the initial field PM_2.5in mainland China are reduced by more than 75%,and the correlation coefficient（CORR）can be improved to 0.95,with more significant improvements in northern and northeastern China.Assimilating only 50%of the PM_2.5 data from ground stations,the RMSE of the initial field PM_2.5 at the validation site in mainland China decreases from64.58μg m^-3 to 47.49μg m^-3,and the CORR increases from 0.63 to 0.85.For the forecast field,assimilating the initial field every 24 hours,the RMSE of PM_2.5 can be reduced by 10%-21%in 24 hours.The RMSE of PM_2.5 can be reduced by 10%-21%within 24 hours,but the assimilation effect is most obvious in the first 12 hours.The assimilation of PM_2.5 can also significantly improve the visibility forecast accuracy.When the PM_2.5 increment is negative,it corresponds to an increase in visibility,while when the PM_2.5 analysis increment is positive,visibility decreases.It is worth noting that the improvement of visibility forecasts by assimilating PM_2.5 is more pronounced during periods of light pollution than during periods of heavy pollution,which are accompanied by low or extreme low visibility,and visibility is more influenced by humidity.（2）Assimilation of humidity data from automatic stations reduces the initial meteorological field specific humidity error to±0.5 g kg^-1,which can provide more accurate humidity background field for the model.The assimilated humidity data improve the model’s relative humidity forecasting capability.By assimilating humidity every 24 hours,the RMSEs of relative humidity forecasts in Beijing-Tianjin-Hebei,Yangtze River Delta,Pearl River Delta,Central China,and Northeast China are reduced by 10.7%,8.2%,5.2%,4.1%and 2.5%,respectively,and the ME are reduced by 8.6%,7.7%,5.1%,3.5%,and 2.3%.The combined assimilation of PM_2.5 and humidity further affect the PM_2.5 and visibility forecasts.When relative humidity increases,PM_2.5concentration increases and visibility is inversely correlated.combined assimilation of PM_2.5 and humidity have a more significant improvement on visibility than assimilation of only one observation alone.The effect of combined assimilated PM_2.5 and humidity on visibility is shown by the overlapping effect of PM_2.5 and humidity assimilation in the first 24 hours,after 24 hours,the effect of assimilated PM_2.5 disappears,and humidity assimilation takes the main role.