Network time synchronization is an important aspect of sensor network operation. It is often achieved by synchronizing the clock of each node in the network to the clock of some reference node. However, it is well known that synchronization error accumulates over multiple hops. This scalability problem presents a challenge for large-scale, multi-hop sensor networks with a large number of nodes distributed over wide areas.

In this thesis we develop the use of spatial averaging as an approach to mitigating the effects of the scalability problem. We first develop a cooperative synchronization technique using spatial averaging that can achieve "perfect" synchronization in the limit of an infinitely dense network. We show that it is possible to maintain a perfect timing signal with equispaced zero-crossings that occur at integer values of the reference time. Second, we study the benefits of cooperative time synchronization using spatial averaging in networks of finite density. We present a protocol that uses spatial averaging to reduce error accumulation in large-scale networks and show that synchronization performance can be significantly improved by increasing network density.