This paper studies distributed subspace tracking in wireless networks based on consensus averaging. Most prior approaches to this require the exchange of many inter-node messages between the arrival of new measurements, forcing communications to happen at a faster timescale than measurements. By contrast, this paper proposes a technique, termed hierarchical subspace tracking, which leverages recent advances in consensus over wireless networks in order to track subspaces when communications and measurements occur on the same timescale. It is shown that the resource consumption of hierarchical subspace tracking scales slowly in the size of the network. Further, it is shown that the convergence speed of hierarchical subspace tracking is the same as if measurements were known globally.