In low-Earth-orbit satellite tracking applications, the apparent wind component due to relative motion between the telescope and the different atmospheric turbulence layers leads to faster turbulence temporal dynamics, and thus increases the servo-lag error of adaptive optics systems. Nevertheless, this apparent wind component can be known a priori, since it only depends on the satellite's orbit. We propose a model-based predictive controller that incorporates a priori wind layer velocity information. The controller uses an autoregressive model computed from layer wind speeds and a C 2 n (z) profile. This approach derives from the spatio-angular linear quadratic Gaussian controller, and benefits from an analytical formulation. We present performance simulation results for two applications: optical communications and satellite observation; we analyse the impact of prediction in terms of relevant performance criteria for each application.