The multiscale phenomena is widely encountered in many nonlinear systems. Uncovering the dynamic characteristics of wind signals is a challenging problem with significant engineering applications. In order to solve this problem, we infer multiscale complex networks from wind speed signals in different seasons and heights. Specifically, we obtain a set of wind speed time series with different time scales using the coarse graining technique and then map those time series into multiscale complex networks using phase space reconstruction technique. Furthermore, we introduce three measures, i.e., clustering coefficient entropy (CCE), mean values of CCE with large scales, variances of CCE, to quantitatively analyze the dynamic and seasonal characteristics of wind signals. The results suggest that the CCE and its derivative index can accurately reflect the dynamic and seasonal changes of wind speed signals. The combination of complex network analysis and multiscale analysis provides an effective method for the detection of wind field behavior.