Evaluation of Numerical Predictions of Boundary Layer Structure during the Lake Michigan Ozone Study

Abstract

The performance of two types of turbulence closures is compared in a 3D numerical investigation of an episode with poor air quality. The first is the Blackadar boundary layer scheme, which has a nonlocal closure for unstable conditions. The second is a 1.5-order scheme, known as Gayno?Seaman (GS), that predicts turbulent kinetic energy and is suitable for simulating foggy as well as dry conditions. In 3D mesoscale simulations of a 5-day air pollution episode over the Midwest, the GS turbulence scheme is found to be effective for reducing model errors in boundary layer depth and surface wind speeds, relative to the Blackadar nonlocal closure. In this case, wind direction and surface temperature simulations have comparable skill with both closures. The 1.5-order GS scheme also is shown to interact well with a four-dimensional data assimilation system that avoids assimilation of smooth analyses below 1500 m. Experiments that combined the 1.5-order boundary layer scheme and a multiscale data assimilation approach produced the lowest model errors overall while producing boundary layer trajectories that are consistent with the observed locations of ozone maxima. The efficiency of the two turbulence schemes was found to be nearly identical, each requiring about 25% of the overall central processing unit computation time.