In this paper, we study the cluster dynamics during network evolution in human mobile network by analyzing the GPS trajectories of taxi collected in two different urban environment. We observe that the size distribution of small clusters exhibits persistence over time, and the behavior of large clusters becomes the key to distinguishing the network state at different time step during the evolution process. We also study the behavior of cluster split and merge process, the results shows that small clusters prefer to take part in split process, and merge process prefers to happen between a pair of clusters which have one large cluster at least. Another interesting result we found is the number of free nodes taking part in join process is approximately equal to those taking part in leave process when the time window is large enough. In the end, we study the lifetime of cluster, and find that if the scale of cluster is larger, then they need lower stability to survive.