This paper studies a cellular network, where base stations move according to traffic/service needs; coverage and capacity offered are adaptive to the time-spatial variations of user demand. In particular, we analyze the performance of a network, where a base station is carried by a UAV (Unmanned Aerial Vehicle, sometimes also denoted as drone). The Unmanned Aerial Base Station (UABS) flies over an urban area, offering coverage and capacity to users that are not served by the Terrestrial Base Stations (TBSs); in our scenario we consider in particular video users with stringent requirements in terms of downlink throughput. The paper proposes an approach to design the trajectory of the drone, that accounts for the actual position of users that are unsatisfied by the TBSs. The approach is based on the identification of clusters made of nearby users to be served, and a simple Nearest Neighbor algorithm. The potential improvements in terms of network capacity (sum throughput) and user satisfaction are estimated, in a scenario where the UABS uses a separate band with respect to the TBSs. This performance increase can be seen as an upper bound to the improvements that would be achieved by using the same band, where the UABS would actually interfere with the TBSs.