Estimating the turbulent flow field of an atmospheric boundary layer from LIDAR data using LES-based 4D-Var

Recent advances in flow field measurement techniques, e.g. LIDAR, allow for a vast amount of information to be collected in the atmospheric boundary layer in the proximity (O(10 km)) of a sensor. This opens up possibilities for new short term turbulent flow field predictions techniques, with time scales up to an hour, which can for example be useful for wind farm power output estimates. The retrieved measurement data is typically spread highly irregular in space and time, due to the typical sweeping motion of the sensors. This makes the estimation of the current flow field state a challenging problem. To this end, we eploy 4D-Var data-assimilation, where we use a large-eddy simulation (LES) model, run on a coarse grid to limit computational complexity, as the sublying model. Virtual LIDAR measurements are taken from a high resolution LES simulation, which is an ideal setup for development and testing purposes. The 4D-Var optimization problem, is solved using L-BFGS, combined with an adjoint LES simulation for the gradient calculation, due to the shear amount of optimization variables. The methodology is illustrated on a pressure driven boundary layer for different case studies.