The conditional entropy of two symbolic sequences which are encoded faithfully from two different time series is minimized at the zero relative shift of the two time series if the two time series have their origins in the same underlying dynamics. This technique is also useful in studying the delayed coupling between different time series. Applications are made to the temperature data derived from five meteorological stations in China, namely Shanghai, Nantong, Changzhou, Nanjing and Hangzhou in order to test the commonality of their underlying dynamics, obtain the relative strength of the dynamics coupling between different variables, and extract the implicit delayed information in the dynamics.