For IOP1, our results suggest that assimilating the wind observations of wind profilers and Aerosond can improve the PBL dynamic structure over central Taiwan. Applying variable error covariance localization is crucial to maintain the impact of wind assimilation and avoid the negative impact on the thermodynamic structure. Due to the significant temperature variation and cold bias, the large innovation occurred easily in the PBL. Assimilating the UAV temperature observation using AOEI alleviates the overly large temperature increment and prevents the unrealistic flow in the analysis, further improving the thermal-direct land-sea breeze simulation. The WRF-LETKF analysis is used to drive the Community Multiscale Air Quality Model (CMAQ) to obtain the air quality simulation. The assimilation of wind data with high vertical resolution corrects the structure of vertical circulation near the terrain and thus better represents the vertical transportation of the pollutants. In addition, the assimilation of UAV temperature corrects the model overcooling at night, which help to suppress the unrealistic strong land breeze and prevent the pollutants from diffusing offshore. In summary, assimilating the PBL observation with the high vertical resolution is beneficial for improving the air-quality simulation related to local emission issues.
The optimized assimilation strategies are further applied to conduct the PBL data assimilation for the IOP3 event. Results with show that assimilating both the PBL wind and temperature data over the complex terrain is important to correctly represent the local circulation (leeside vortex) and the transport of air pollutants.

