Mixing layer height (MLH) is retrieved at an urban site in Seoul, Korea, from 2-year collocated remote sensing instrument measurements: elastic aerosol lidar, ceilometer, and wind Doppler lidar (WDL). All three measurements captured clear diurnal patterns of MLH development and break down, but large discrepancies were noted in the nocturnal MLH and the onset time of MLH development. Vertical wind speed standard deviation (σw) from WDL measurements displayed low MLHs during nighttime, compared to MLH retrieved from aerosol backscattering intensity profiles (mean bias = −0.21 km). To this extent, mixing represented by σw showed close correlation with surface heating but significant discrepancies with the vertical distribution of atmospheric constituents, especially at night. Comparison of MLH from elastic aerosol lidar and WDL with WRF-Chem simulation results showed that WRF-Chem MLH closely resembled MLH retrieved from WDL, indicating that WRF-Chem sufficiently simulated turbulence, but displayed lower nocturnal MLH compared to lidar-derived MLH. Furthermore, investigation of the vertical profiles of PM2.5 in the model displayed confined distribution of aerosols within the nocturnal MLH, leading to overestimation of surface PM2.5 concentration. For the case studies of May 12 and 18, 2016, implicating full mixing within the MLH defined by aerosol lidar measurements on WRF-Chem PM2.5 vertical distribution simulations was shown to significantly reduce WRF-Chem PM2.5 simulation errors from mean normalized bias of 52% to 19%.