This paper studies dynamic data communications between airplanes and a control tower, where the control tower needs to monitor the state of each aircraft close to the airport or on the ground in real time. Given limited communication bandwidth, it is impossible for the control tower to communicate with all aircrafts at the same time. This paper focuses on the problem of optimal scheduling of data communications for the control tower to acquire information from aircrafts to minimize tracking errors. A dynamic learning problem with limited communication bandwidth is formulated in this paper where the objective is to minimize the total variance of real-time tracking. To solve the problem, a dynamic scheduling algorithm for data communications is proposed, which prioritizes data communications based on the tracking variances of the aircrafts, channel conditions and importance of the information. Our simulations demonstrate that our algorithm outperforms policies such as a round robin policy.