Four-dimensional Data Assimilation of Atmospheric Boundary Layer Using an Adjoint Model

The four-dimensional variational data assimilation technique is used to recover complete three-dimensional velocity and temperature fields of a convective boundary layer flow generated by Large-eddy Simulation. The "simulated" observation data contains three-dimensional radial velocity fields and a mean virtual potential temperature profile. We test the sensitivity of retrieval quality to the observation frequency, the eddy viscosity and diffusivity, and to the temporal and spatial smoothness penalty functions. An experiment using the optimal parameters derived above is then conducted. The retrieved velocity field shows good agreement with the corresponding LES data, but the retrieved temperature is less accurate.