Wind turbines operating in the wake of other turbines experience a strongly altered inflow which can yield a reduced power output and increased loads on a turbine. Good wake models are therefore crucial for the optimization of wind farm layouts and the controlling of wind farms. In this work, we aim for developing simplified wake models which include dynamical aspects of the wake but avoid too complex numerical simulations. For this purpose, we decompose the wake flow into a superposition of spatial modes to obtain a reduced order description of the wake. For data from complex numerical simulations (large eddy simulations), it could be shown that relevant aspects of the wake could be captured using only a few modes. To model the temporal dynamics of the wake, we plan to model the weighting coefficients of the obtained modes using stochastic differential equations.