Considering Background PBL Structure Information In Surface Observation Assimilation In The GRAPES Mesoscale Model

With the advantages of dense site distribution,high observation frequency and strong realtime data,the surface observation data can reflect the dynamic and thermal changes of the atmosphere boundary layer,and it has high value in high-resolution model.But the assimila t io n utilization of surface observation in current model is very low,and it has not been able to effectively enter the numerical model to improve the numerical prediction results,resulting in waste of resources.Effective assimilation of surface observations into a mesoscale threedimensional variational data assimilation(3DVar)scheme requires spreading surface observations upward with anisotropic and inhomogeneous error covariance structures,but 3DVar analysis assumes a relatively constant vertical correlation of background error.To overcome this weakness of 3DVar in assimilating surface observations,we adopted the technique proposed by Benjamin(2004)of using the background planetary boundary layer(PBL)structure information and applied it to the Chinese GRAPES mesoscale 3DVar data assimilation system.However,the short-term forecast of the background PBL structure predicted by the numerical model may be inconsistent with actual observations.Therefore,we utilized high vertical resolution(~9.6m)radiosonde data and mesoscale model background data to analyze the difference between the predicted background PBL structure and the observed PBL structure under stable and unstable conditions during the diurnal cycle.The results showed that the vertical autocorrelation of innovations of the surface potential temperature from radiosonde observations is highly dependent on both the background and observed PBL structures under stable and unstable conditions.When both the background and observed PBL structures are either stable or unstable,the vertical autocorrelation from surface upward reaches 2000 m.However,when the background and observed PBL structures are inconsistent,the vertical autocorrelation from surface upward decreases to 600 m.This indicates that Benjamin's PBL-depth assimilation technique,which depends entirely on the background PBL structure,should be modified to include both the background PBL structure and the observed PBL structure information in the assimilation process.Therefore,Benjamin's assimilation technique based on background field structure is adjusted according to the statistical results.The adjustment contents mainly include three aspects:(1)replacing the temperature variable and adjusting the fitting height interval;(2)setting of the vertical layer observation error;(3)setting of assimilation height.In addition,a heavy precipitation process under stable PBL condition on 7 May 2017 is selected to analys is the performance of the modified PBL-depth assimilation technique,the results show that comparing with traditional 2m surface temperature analysis,PBL-depth assimilation technique conserves the PBL structure and a small improvement in the precipitation BIAS threshold score is achieved.