As a distributed system, most applications in wireless sensor networks (WSNs) require clock synchronization as support. However, clock crystal oscillators in WSNs are highly susceptible to the environment. The problem of clock synchronization among sensors in WSNs is urgently to be solved. Therefore, in this paper, a state-space model considering the distributed characteristics of WSNs is set up to solve clock synchronization problems. Based on this model, a diffusion Kalman filter is put forward to implement compensation of clock offset and clock skew. Furthermore, an event-based diffusion Kalman filter algorithm is proposed, in which, the eb-DKF will be executed only when some event happens. Through this proposed solution, it can achieve close performances with DKF implement but reduce computational cost significantly, which is particularly adaptive to the low power and low cost properties of WSNs. At last, experiments is shown to support these theoretical algorithm well.